Providers and deployers of AI systems: You only have a couple of weeks left until you must explicitly inform users when they are interacting with AI content.

To assist in the effort, the European Commission (Commission) has published guidelines to help AI deployers get in line with the AI Act’s transparency obligations, which will begin to go into effect on August 2.

After that, companies providing AI systems must alert users when they are interacting with AI. They must also tell users when they have been exposed to deepfakes, “emotion recognition,” or biometric categorization systems, or when they are given AI-manipulated content in matters of “public interests without human review or editorial control.”

Henna Virkkunen, the Commission’s executive VP for tech sovereignty, security and democracy, said in a statement, “with today’s guidelines, the Commission supports the smooth and effective application of the AI Act to make AI systems interacting with people such as chatbots and AI agents and AI content more transparent and trustworthy. These guidelines support providers and deployers in meeting their obligations under the AI Act, while helping citizens know when they are interacting with AI.”

Systems must include machine-readable markers to reveal such content, to reduce “the risk of deception and manipulation” and build public trust in AI.

“Generative systems have collapsed the cost of producing convincing content while the cost of judging it stands where it always stood,” said Sanchit Vir Gogia, chief analyst at Greyhound Research. This requirement is “an attempt to restore friction to that imbalance.”

A company’s non-compliance could result in fines anywhere from €750K (about $856K) to €15M (about $17 million), or even up to 3% of its total worldwide annual revenue.

Transparency requirements

The EU AI Act’s transparency requirements apply to “natural or legal persons,” public authorities, agencies, or other bodies that develop AI systems, or have them developed, and place them on the EU market or into use under their name or trademark. This means all companies, regardless of whether or not they are EU-based.

“Systems placed on the European market, put into service there, or producing outputs used there are inside the field, wherever the developer sits,” Gogia noted.

Applicable systems must be intended to interact directly with “natural persons”; these systems include AI-enabled chatbots or conversational agents, AI companions, or coding agents. However, AI-enabled tools like recommender systems, spam filters, authentication, search and retrieval, transcription, text and code auto-completion, or predictive maintenance do not fall under the rule.

Specific outputs such as AI-generated text, images, video, and audio must contain a machine-readable mark. Deepfakes and public interest-related text created by AI without human review or control must be clearly labeled, however, deepfake content that is “artistic, creative, satirical, or fictional” is largely exempt.

AI content must be marked with one of three labels: “AI,” “Fully AI-generated,” or “Partially AI-modified.” For instance, “Fully AI-generated” applies when news summaries, music, art, or videos have been created without any human oversight (apart from prompting), while “partially AI-modified” could mean a person’s face is swapped into an authentic photograph to create a deepfake.

The three icons are publicly available for free use; enterprises can download zip files in PNG and SVG formats.

Most of the Act’s transparency rules begin to go into effect on August 2. But AI systems placed on the market before then will have some leeway; they must be in compliance by December 2.

However, a four-month allowance “on one obligation, for one population of systems, contingent on one procedural step, is not a strategy,” Gogia emphasized. Enterprises should plan to comply by August 2 and “treat any relief that arrives as margin.”

A consistent code of practice

Along with the transparency guidelines, the Commission has introduced a code of practice that essentially serves as a gesture of good faith. When signed, it can provide “legal certainty” and a “simple and practical” way to demonstrate compliance with the AI Act, according to the Commission. Signatories can also collaborate through the ‘Signatory Taskforce,’ which will share practices and advance technologies around marking and labeling practices.

Providers that choose not to sign must comply through other methods and demonstrate that those methods are “adequate” through assessment by surveillance authorities, according to the Commission.

Non-signatories “keep their flexibility, and will face more case-by-case scrutiny for it,” said Gogia.

Criteria for compliance

Shashi Bellamkonda, principal research director at Info-Tech Research Group, pointed out that the transparency requirements apply to content only when three criteria are met: It has been published, is informative to the public, or is on matters of public interest.

B2B business content or blogs may not need an AI disclosure if they do not meet these criteria, he noted. Also, published text that has undergone human review or is under editorial control does not need to be labeled. Editorial control means that a person must hold the ultimate legal responsibility for the publication of the content.

Many companies like Google, Adobe, and LinkedIn have already established ways to identify images marked as AI-generated. Meta has made it a requirement, but the creator has to add the AI-generated label, Bellamkonda said.

“This is a good move for guardrails around public information, and companies with good compliance and ethical oversight may not have to worry about this,” he noted. But as a general practice, companies should disclose AI-generated content and state whether it has been human reviewed.

Creating a transparency pipeline

Establishing full transparency means identifying who carries the responsibility for the content, whether the marking survives real use, not just testing, and what evidence will defend the decision, Gogia said.

Concerns cluster around responsibility, durability and evidence. Several organizations usually touch one piece of content, and none controls the whole chain, which is why contracts become the “pressure point,” he said. Most current agreements were written to deliver software and say “almost nothing” about provenance persistence, verification access, or evidence retention.

The durability concern is the most difficult, Gogia noted, because marking performs well in controlled settings but “badly in ordinary life.” Meta, for one, said its invisible watermark was designed to survive cropping; a published test, however, found the company’s preview detector missed 55% of cropped images.

“CIOs should ask which platform can actually provide evidence before believing its dashboard,” said Gogia.

Disclosure of AI use must be “clear, distinguishable and accessible,” he emphasized. “A notice buried in lengthy terms, or reachable only through determined clicking, satisfies nobody, least of all a market surveillance authority.”

Sustained compliance is a “living control” requiring a central record of systems, duties and evidence; testing taking place where the user meets the control rather than where the developer built it; and continuous supplier assurance. Enforcement will vary by country, so keep one common baseline with local overlays, Gogia said.

His advice: Inventory every system that talks to people, generates content, or gauges sentiment; classify provider and deployer roles; place disclosures at first interaction; define substantive human review; keep the evidence.

Marks and provenance signals should be tested after content undergoes cropping, compression, translation, transcription, and other editing, Gogia said. A useful audit starts from a real output and follows its “pulse” through generation, editing and publication, identifying at “each beat” the responsible party, the surviving mark, and evidence for exceptions. Missed labels should also be traced for root cause and recurrence.

To ensure compliance, before August 2, enterprises need a prioritized inventory, live disclosures on the highest-risk use cases, and a “named owner for every control,” he noted. In the first 30 days, they should stabilize and test; in the first 90 days, push requirements into procurement processes as a standing discipline. Procurement must secure commitments on marking methods, known failure modes, and evidence access, with explicit notice if/when any of them change.

“The sensible architecture is a common transparency baseline carrying traceability, responsibility, and evidence, with jurisdictional overlays for language, sector rules, and local practice,” Gogia said.

This article originally appeared on CIO.com.

As Apple gears up to anoint John Ternus the new company CEO in September (while current leader Tim Cook takes a seat on the board) the company appears to be firing on all cylinders ahead of the leadership transition. 

What’s going well

Just look at the evidence: 

• Apple is building market share across its entire product range; even memory-driven price inflation doesn’t seem to have dampened demand for its hardware yet.
• While Apple had to raise prices, the company’s MacBook Neo remains seriously popular. It’s sitting atop Amazon’s US best-selling chart, which currently includes six Macs in the top 10. The Neo has topped this chart since its introduction.
• Apple’s iPhone 17 series continues to sell well, with recent market data showing sustained growth. Both Counterpoint and IDC tell us that iPhone shipments continue to increase, even as other vendor shipments slide.
• IDC analyst Francisco Jeronimo recently estimated that Apple’s upcoming foldable iPhone Ultra could grab 29.4% of global folding smartphone sales this year, rising to 34.9% in 2027.
• The company’s new 27 series of operating systems is attracting a great response as beta testers report that it is already solid, stable, and performing well.
• The AI narrative has really changed, with analysts no longer quite so starry-eyed at the prospects for the big frontier AI firms. Apple’s edge-AI-enabling approach is winning converts.

What’s coming up

The company’s newly-filed lawsuit against OpenAI may or may not succeed, but it will certainly help consolidate recognition of the importance of Apple’s designs and intellectual property in whatever hardware emerges from the AI firm. It also means both Apple and OpenAI are already competing in hardware, even though neither company yet offers anything that directly challenges the other. 

Apple has just set out its stall to brand-loyal fans in a big way and did so before OpenAI gets to woo the same set of customers with a wriggle of its Jony Ive-tinged talisman.

The stage is set for intense competition between the two. Though some say Apple’s needs to improve employee retention, if it does find proof of efforts to use recruitment to engage in industrial espionage, it’ll be easier to represent its own products as being the OG for new hardware. 

If nothing else, it means consumers will forever be asking, “If OpenAI’s designers are so good, why did it need to poach them from Apple?” Doubt is a weapon.

Managing perception

It doesn’t matter how the case goes, because there fight is already affecting consumer psychology. It also means that as Ternus prepares to take his seat atop the rainbow-colored Apple throne, we can already size him up. “A man is measured by his enemies,” Joe Abercrombie wrote in “The Trouble With Peace.”

Given the proximity of the leadership transition, it’s highly probable that Ternus signed-off on the litigation; in doing so he — and Apple — tell us to expect more of the same. 

Apple has, rightly or wrongly, decided that OpenAI will become its new existential bugbear, following in the footsteps of Microsoft Windows, Real Networks, Adobe Flash, Android, and Samsung, all of whom have been useful foils against which Apple has been able to build and maintain its identity.

Looking at that list, you’d be tempted to believe that nothing much is new. Apple has often defined itself by the enemies it sometimes keeps. What has been will be again, which in this case means even as OpenAI attempts to carve out an identity as a hardware manufacturer delivering solutions to compete with Apple and Google, Ternus’ team’s looks to drive a consensus-shaped wedge into the pro-LLM propaganda. 

That blow comes as Apple finally gets its act together around AI, and as the company prepares for a future in which the world’s most-used wearable device also becomes the wearable way to woo Siri AI.

My kingdom come

Rising market share, powerful solutions, an increasingly recognized and respected approach to AI, and an ideological crusade — these details constitute Apple’s place today and are Tim Cook’s coronation gift to Ternus. He’s passing along a strong and hyper-profitable baton that screams of timeliness and relevance even as the company gets set, ready, to go with a year or two of new product designs, new product families, and a 20^thanniversary iPhone.

This is Apple’s party. OpenAI’s name didn’t make the list. 

You can follow me on social media! Join me on BlueSky,  LinkedIn, Mastodon and subscribe to The Core.

Among a range of factors leading to a wave of tech sector layoffs in 2026 is the rapid rise of artificial intelligence and automation. Companies are reconfiguring their workforces to leverage AI for increased efficiency and reduced operating costs. This realignment and reduction is implemented even by companies reporting strong financial performance.

But it’s not just AI leading to workforce cuts. Complementing this technological shift are ongoing economic uncertainty, inflation, and higher interest rates, compounded by a chip shortage and rising energy costs. This mix is driving companies to cut costs and streamline operations for increased efficiency.

According to data compiled by Layoffs.fyi, an online tracker that keep tabs on job losses in the technology sector, 123,941 tech employees were laid off at 269 companies in 2025. The site also reports that 71,981 government employees were laid off by DOGE alone, with 182,528 total federal workers laid off.

Here is a list — to be updated regularly — of some of the most prominent technology layoffs the industry has experienced recently.

Notable tech layoffs in 2026

• Monday.com
• Microsoft
• Meta
• Cisco
• Cloudflare
• Oracle
• Atlassian
• Salesforce
• Amazon
• Ericsson

July 22, 2026: Monday.com cuts 20% of its workforce to restructure for the AI era

The company says the decision to cut 620 jobs isn’t about margins, but about creating a flatter organization built around AI agents, autonomous teams, and deeper customer engagement.

July 6, 2026: Microsoft cuts 4,800 jobs, primarily in sales and Xbox teams

As the company trims thousands of jobs, it’s also investing in embedded engineering teams and AI infrastructure. The layoffs come several weeks after the company offered 8,750 US employees voluntary retirement buyouts.

June 5, 2026: Tech industry cut 38,242 jobs in May, worst since 2024

AI was blamed for 40% of the job cuts in May, up from 7% in January, according to research by employment placement company Challenger, Gray & Christmas.

May 20, 2026: Meta cuts 8,000 jobs, around 10% of workforce

The cuts are expected to expected to hit Meta’s engineering and product teams the hardest, arriving as Meta pivots toward AI to boost efficiency across its organization, according to Yahoo Tech.

May 13, 2026: Cisco to cut nearly 4,000 jobs despite strong growth in AI, enterprise networking

Despite reporting positive financial news — including record third-quarter revenue of $15.8 billion, a 12% year-over-year increase — Cisco said it will eliminate almost 4,000 jobs.

May 7, 2026: Cloudflare to cut 1,100 jobs in AI-focused restructuring

About 20% of Cloudflare’s global workforce will be culled as the company pivots for the agentic AI era, Reuters reported.

April 1, 2026: Oracle to cut up to 30,000 jobs globally, putting enterprise support and roadmaps at risk

Oracle began laying off employees on March 31 in what could be the largest workforce reduction in the company’s history. Employees received termination emails at 6 a.m. local time with immediate system lockouts and no prior warning. (Note: in June, CNBC put the final layoff tally at 21,000.)

March 12, 2026: Atlassian cuts 1,600 jobs to fund AI and enterprise expansion

Atlassian will reduce its global workforce by approximately 10%, eliminating around 1,600 roles, as the collaboration software maker redirects capital toward artificial intelligence development and enterprise sales.

March 11, 2026: Tech layoffs surpass 45,000 in early 2026

A recent analysis by RationalFX found 45,363 job cuts globally so far this year—with roughly 68% or more than 30,000 occurring in the U.S. — highlighting ongoing workforce cuts even as many tech companies report strong revenue growth.

February 10, 2026: Salesforce lays off staffers as executive leadership churn continues

Salesforce has reduced close to 1,000 roles earlier this month across teams, including marketing, product management, data analytics, and its Agentforce AI unit, Business Insider reported, quoting employees familiar with the matter.

January 23, 2026: Amazon layoffs expected to disproportionately hit AWS and tech talent

As the market slows down, AWS and other Amazon units are preparing for another round of layoffs, which is expected to overwhelmingly impact tech talent. An email from HR leader Beth Galetti on Jan. 28 confirmed 16,000 job cuts.

January 15, 2026: Ericsson plans to shed 1,600 jobs in Sweden

 Ericsson lans to cut some 1,600 jobs in Sweden, the telecommunications equipment maker said doubling down on recent cost-saving measures that have helped it weather a prolonged downturn in telecoms spending, Reuters reports.

January 13, 2026: Meta plans to cut around 10% of employees in Reality Labs business

Meta plans to cut around 10% of the employees in its Reality Labs division who work on products including the metaverse, according to three people with knowledge of the discussions, according to The New York Times.

Layoffs in 2025

• Cisco
• Oracle
• Windsurf
• Intel
• Microsoft
• Crowdstrike
• HPE
• Autodesk
• HPE
• CISA
• Workday
• Salesforce
• Meta

Global tech-sector layoffs surpass 244,000 in 2025

Economic uncertainty, elevated interest rates, and AI adoption have driven workforce reductions across tech companies worldwide, according to a RationalFX report.

October 28, 2025: Amazon to cut 14,000 jobs across company

Amazon will reduce its overall workforce by 14,000, cutting layers of management across the company and hiring in some areas to support its “biggest bets”.

August 18, 2025: Cisco and Oracle to cut hundreds of Bay Area jobs

Tech companies Cisco and Oracle are cutting hundreds of jobs across the Bay Area. Cisco will eliminate 221 positions at its Milpitas and San Francisco offices, effective Oct. 13. Oracle is reducing 101 positions in Santa Clara on the same date 

August 5, 2025: 3 weeks after acquiring Windsurf, Cognition offers staff the exit door

Cognition, the AI coding startup that acquired rival company Windsurf three weeks ago, laid off 30 employees last week and is offering buyouts to the roughly 200 remaining employees on the team, reports The Information.

July 25, 2025, Intel to lay off 22% of workforce, CEO Tan signals ‘no more blank checks’

Intel will reduce its workforce to 75,000 employees by the end of 2025 as new CEO Lip-Bu Tan implements sweeping changes designed to transform the struggling chipmaker

July 8, 2025, Intel layoffs begin: Chipmaker is cutting many thousands of jobs

Intel has begun laying off employees across the company. CEO Lip-Bu Tan told workers back in April to expect major layoffs at Intel in the coming months as the chipmaker slashes costs and overhauls its organization after years of technical setbacks and falling sales.

July 2, 2025: Microsoft will cut 9,000 workers

Microsoft will lay off about 9,000 employees, a source familiar with the workforce cut told CNBC.  The cuts will reportedly affect less than 4% of Microsoft’s global workforce and will impact different teams, geographies and levels of experience. This is the latest in a string of cuts the tech giant has made this year.

June 17, 2025: Intel looks to factory layoffs to return to profitability

Intel will lay off up to 20% of its manufacturing sector employees starting in July, according to media reports, as the company looks for options as it seeks a return to profitability. The cuts reportedly will be made around the world, but some of the layoffs will be closer to home, according to a report in The Oregonian citing an internal company memo from Intel manufacturing Vice President Naga Chandrasekaran.

May 7, 2025: CrowdStrike to lay off 5% of staff

CrowdStrike announced a plan to cut about 500 roles, roughly 5% of its workforce, to streamline operations and reduce costs. The cybersecurity company will incur about $36 million to $53 million in charges related to the layoffs

March 6, 2025: HPE cuts 2,500 jobs, remains committed to Juniper buy

CEO Antonio Neri told Wall Street analysts that HPE would begin implementing a cost-cutting program involving layoffs of about 2,500 employees over the next 18 months. HPE employs about 61,000 people worldwide.

Feb. 27, 2025: Autodesk to lay off 9% of workforce

Software maker Autodesk is laying off 1,350 staff. With the rise of subscription and multi-year contracts billed annually, and self-service enablement, it finds it needs fewer sales staff, CEO Andrew Anagnost said in a message to employees. And with its cloud, platform, and AI products proving most profitable, it’s concentrating its staff and investments there.

Feb. 27, 2025: HP to lay off 2,000 more

As part of an ongoing restructuring, HP plans to lay off up to another 2,000 workers. In recent weeks, the company has tried — unsuccessfully — to do away with telephone support staff by forcing callers to wait for at least 15 minutes if they refuse to use self-service support resources online. The company swiftly backtracked, but wider job cuts are still on.

Feb. 21, 2025: CISA lays off 130

Government employees get laid off too: In this case, 130 workers at the US Cybersecurity and Infrastructure Security Agency are being shown the door as a result of a DOGE decision. Cybersecurity experts are concerned that the cuts will harm the international collaborations that CISA has fostered, quite apart from their concerns about the security of the DOGE layoff process itself.

Feb. 5, 2025: Workday lays off 1,750

As it moves to invest more in AI and international growth, Workday is laying off 8.5% of its workforce and disposing of unused office space. Some analysts fear the cutbacks will affect the company’s customer service — unless AI can pick up the slack.

Feb. 4, 2025: Salesforce lays off over 1,000

At the same time as it’s hiring sales staff for its new artificial intelligence products, Salesforce is laying off over 1,000 workers across the company, according to Bloomberg. As of June, 2024, the company had over 72,000 employees, according to its website. Salesforce did not comment on the report. In 2024 the company reportedly laid off around 1,000 staff too, in two waves: January and July.

Jan. 14, 2025: Meta will lay off 5% of workforce

Mark Zuckerberg told Meta employees he intended to “move out the low performers faster” in an internal memo reported by Bloomberg. The memo announced that the company will lay off 5% of its staff, or around 3,600 staff, beginning Feb. 10. The company had already reduced its headcount by 5% in 2024 through natural attrition, the memo said. Among those leaving the company will be staff previously responsible for fact checking of posts on its social media platforms in the US, as the company begins relying on its users to police content.

Tech layoffs in 2024

• Equinix
• AMD
• Freshworks
• Cisco
• General Motors
• Intel
• OpenText
• Microsoft
• AWS
• Dell

Nov. 26, 2024: Equinix to cut 3% of staff

Despite intense demand for its data center capacity, Equinix is planning to lay off 3% of its workforce, or around 400 employees. The announcement followed the appointment of Adaire Fox-Martin to replace Charles Meyers as CEO and the departures of two other senior executives, CIO Milind Wagle and CISO Michael Montoya.

Nov. 13, 2024: AMD to cut 4% of workforce

AMD will lay off around 1,000 employees as it pivots towards developing AI-focused chips, it said. The move came as a surprise to staff, as the company also reported strong quarterly earnings.

Nov. 7, 2024: Freshworks lays off 660

Enterprise software vendor Freshworks laid off around 660 staff, or around 13% of its headcount, despite reporting increased revenue and profits in its fourth fiscal quarter. The company described the layoffs as a realignment of its global workforce.

Sept. 17, 2024: Cisco lays off 6,000

After laying off around 4,200 staff in February, Cisco is at it again, laying off another 6,000 or around 7% of its workforce. Among the divisions affected were its threat intelligence unit, Talos Security.

Aug. 20, 2024: General Motors lays off 1,000 software staff

More than 1,000 software and services staff are on the way out at General Motors, signalling that it could be rethinking its digital transformation strategy. In an internal memo, the company said that it was moving resources to its highest-priority work and flattening hierarchies.

August 1, 2024: Intel removes 15,000 roles

Intel plans to cut its workforce by around 15% to reduce costs after a disastrous second quarter. Revenue for the three months to June 29 stagnated at around $12.8 billion, but net income fell 85% to $83 million, prompting CEO Pat Gelsinger to bring forward a company-wide meeting in order to announce that 15,000 staff would lose their jobs. “This is an incredibly hard day for Intel as we are making some of the most consequential changes in our company’s history,” Gelsinger wrote in an email to staff, continuing: “Our revenues have not grown as expected — and we’ve yet to fully benefit from powerful trends, like AI. Our costs are too high, our margins are too low. We need bolder actions to address both — particularly given our financial results and outlook for the second half of 2024, which is tougher than previously expected.”

July 4, 2024: OpenText to lay off 1,200

OpenText said it will lay off 1,200 staff, or about 1.7% of its workforce, in a bid to save around $100 million annually. It plans to hire new sales and engineering staff in other areas in 2025, it said.

June 4, 2024: Microsoft lays off staff in Azure division

Microsoft laid off staff in several teams supporting its cloud services, including Azure for Operations and Mission Engineering. The company didn’t say exactly how many staff were leaving.

April 4, 2024: Amazon downsizes AWS in a fresh cost-cutting round

Amazon announced hundreds of layoffs in the sales and marketing teams of its AWS cloud services division — and also in the technology development teams for its physical retail stores, as it stepped back from efforts to generalize the “Just Walk Out” technology built for its Amazon Fresh grocery stores.

April 1, 2024: Dell acknowledges 13,000 job cuts

Dell Technologies’ latest 10K filing with the US Securities and Exchange Commission disclosed that the company had laid off 13,000 employees over the course of the 2023 fiscal year; it characterized the layoffs and other reorganizational moves as cost-cutting measures. “These actions resulted in a reduction in our overall headcount,” the company said. A comparison to the previous year’s 10K filing, performed by The Register, found that Dell employed 133,000 people at that point, compared to 120,000 as of February 2024. Dell announced layoffs of 6,650 staffers on Feb. 6, but it is unclear whether those cuts were reflected in the numbers from this year’s 10K statement.

See news of earlier layoffs.

Traditional phishing techniques are in decline as a result of the disruption of the Tycoon2FA phishing-as-a-service (PHaaS) platform, Microsoft said in a new report, “Email threat landscape: Q2 2026 trends and insights”.

“Phishing volume linked to the platform fell 92% from pre-disruption averages, including QR code phishing and CAPTCHA-gated phishing both declining from their March highs,” the company wrote in the report.

The takedown reduced activity across multiple phishing categories, forcing attackers to shift to newer delivery methods.

Riding this shift in were a few notable phishing campaigns, including an automated business email compromise (BEC) campaign that reached 42,000 organizations in under three hours, and a multi-stage phishing campaign that used nested email (EML) files, calendar invitations, and a Microsoft authentication redirect to deliver malware.

To counter phishing attacks, Microsoft recommends blocking emails containing known bad URLs/ subject fields, enabling password-less authentication methods, or moving to MFA for accounts that still require passwords.

Tycoon2FA disruption sent attackers exploring

The take-down of Tycoon2FA forced its operators to abandon portions of their infrastructure and rework hosting, domain registrations, and delivery mechanisms.

“After falling 15% in March and another 22% in April, Tycoon2FA-linked phishing volume dropped 74% in May to just 1.5 million messages, then fell another 20% in June to 1.2 million, by far the lowest monthly volumes observed in at least a year,” Microsoft said.

The decline extended to QR Code lures and fake CAPTCHA pages, two phishing techniques in which Tycoon2FA accounted for 12% and 14% of industry activity in June, respectively. This indicated that the platform’s customer base had not been able to migrate to a replacement infrastructure.

But cutting off one head of the hacker hydra only gave rise to new tactics elsewhere.

The adaptation came in the form of using Microsoft Teams as a social engineering channel. Attackers established conversations to build trust before attempting credential theft or delivering malicious payloads. “Teams-based phishing volume climbed steadily throughout Q2, with the average number of detected attacks rising 19% from March to April, holding roughly flat into May (+1%), then increasing another 10% into June,” Microsoft said.

Microsoft also observed a highly automated BEC campaign that reached over 67,000 users using scripted emails, Amazon Simple Email Service (SES), and engagement tracking, alongside a separate phishing campaign targeting 107,000 users that abused Microsoft’s authentication flow and trusted cloud services, including Teams archive recording and ICS calendar invite, to disguise malware delivery behind legitimate infrastructure.

Phishing changes but the defense doesn’t

While QR Code and Captcha-based phishing attacks dropped significantly in the second quarter, business email compromise (BEC) charted jumped 121% between March and April, before dropping down again in May.

QR Code phishing represented 8.3 million attacks in June 2026, down from a peak of 18.7 million in March. Similarly, Captcha-gated phishing fell from 12 million attacks in March to 2.2 million in June.

BEC attacks hit 9 million in March, falling to 3.9 million in June.

But even as these phishing classics lost momentum and newer techniques emerged, Microsoft’s defensive advice remained rooted in the basics. It noted organizations should complement email filtering with phishing-resistant authentication such as passkeys and phishing-resistant MFA to reduce the effectiveness of credential theft campaigns.

The company also recommended strengthening Exchange Online Protection and Microsoft Defender for Office 365 with capabilities such as Safe links and Zero-hour Auto Purge (ZAP), in which malicious emails already delivered to mailboxes are removed before they are read, alongside enforcing password-less authentication methods like Windows Hello, FIDO keys, and Microsoft Authenticator.

Microsoft concluded its report with a list of indicators of compromise (IoCs) from the threats observed in the quarter to support detection efforts.

This article first appeared on CSO.

The European Commission has fined Google a total of €890 million ($1 billion) for its breaches of the Digital Market Act (DMA).

Just over half the fine — €460 million — was because Google illegally gave preference to its own services in Google Search results.

The remainder was because in the Google Play store for Android apps, the company prevented app developers from leading consumers to alternative, often cheaper, purchase channels. Under the DMA, app developers who distribute their apps via Google Play or Apple’s App Store should be able to inform customers of alternative offers.

Now Google must give third-party services featuring in its results the same treatment as its own services, and allow developers of apps in the Play Store to communicate about offers both in and outside the Play Store, or face further fines.

The Commission first raised these issues with Google in March 2025. In April of this year, the Commission laid out plans as to how Google should allow other third-parties to share its searches, suggestions that the tech firm firmly resisted. Earlier this month, the Commission also said  Android should be open to other AI agents and not limited to Google’s own Gemini.

Google is not the only US company to have fallen foul of the DMA. In April 2025, Apple was fined €500 million for breaching the Act and, last month, the Commission fired the first shots at cloud hyperscalers Microsoft and Amazon.

When it comes to building an effective AI stack, context is king and power isn’t everything it’s cracked up to be.

“Smaller, smarter, safer — this is a bet our company has taken in how we deploy AI internally,” said Ricky Thakrar, head of sales and account management at Zoho, provider of a suite of popular cloud-based software solutions for sales, marketing, and finance.

“I’m on the business side, and so decisions made by our CIO and IT folks affect me directly, and my teams’ workflows and processes,” he added.

Speaking to a room of tech leaders at the CIO 100 Leadership Live New York event last week, Thakrar explained that every company wants the speed of AI-generated work wedded to the quality of human work, even though these two are diametrically opposed. No amount of model upgrades or spend will close that gap, so the only way forward is to architect your way out. Thakrar encapsulated this idea in a simple formula:

• Smaller: Stop deploying maximum firepower on every task. Many tasks don’t need it.
• Smarter: The system around the model decides more than the model does.
• Safer: Verify at the point a mistake gets locked in, not just downstream of it.

He noted that organizations that win with AI won’t be those deploying the biggest, most powerful models or the most sophisticated architecture, but the ones that figure out that the model is the easy part and the right architecture is harder. That means understanding the hardest element, and the biggest differentiator, is building a human system that learns and compounds alongside agentic systems.

To get it right, organizations need to prioritize the context layer. The size of frontier models like the GPT series, Claude, and Gemini mostly exist to compensate for missing context, Thakrar explained. Without enough context, models need to be able to reason harder and infer more about what a user actually means because it doesn’t know the user’s account, process, or history. A rich context layer makes it possible for enterprises to run workloads on much smaller, lower-power models.

“The intelligence moves from the model into the architecture around it,” he said.

A steep learning curve

One of Zoho’s earliest AI agents was a churn management agent to help the account management team detect churn in customer subscriptions. So when a subscription became inactive, the agent would collect context from notes, meeting recordings, and Zoho’s data enrichment tool, then create a summary of reasons the account might have churned, and schedule a call.

“What happened was I got this churn agent a couple months later, already embedded in our CRM, and within a week my team no longer trusted that agent,” Thakrar said. “The reason is we forgot to collect one very key point.”

In Zoho’s CRM, when a customer buys a bundle of products, that bundle is represented as a single line item. That means the status of any products the customer may have previously purchased individually changes to inactive as they’re moved to the bundle. That’s not churn, but it was interpreted it that way. Zoho fixed it in the second version of the agent.

Then a new problem arose. Many potential customers first purchase Zoho products as pilots or sandboxes. As those customers move from pilot to live instance, they close down the pilot versions. And again, the CRM would record that as subscriptions going inactive.

“The trust deteriorates again because everyone got excited for version 2,” Thakrar said.

Sometimes, a certain product might not be the best fit for a customer and Thakrar’s team will suggest the customer move to another product. That’s deliberate churn, not a churn risk.

“You may have a similar story like this where the agent sounds so good, it’s going to do something quick and add value, but it’s missing context from the account managers, and there are so many more pieces we’re still building out,” Thakrar said. “It’s been almost a year and the problem I have is my team still doesn’t trust it. They’ll see [a message from the agent] and go out and do all the research anyway to make sure it gave the correct answer.”

The team is more on top of potential churn, though, but the promised productivity gains have yet to materialize because the agent has to earn back lost trust due to a lack of context.

“My goal for this year is having an AI-assisted customer journey from sales to account management where the handoff is clean, the context flows, and every piece of information we gather about a customer is weighed, identified, and coached so the sales team can close more deals,” he said.

Zoho’s early experience with agents has led to the idea that constrained, context-rich, deterministic architectures consistently outperform expensive models bolted onto fragmented systems. It all comes down to three pillars: routing, harness, and specialization.

Routing

Routing is about sending workloads to the proper model for the job, which entails providing enough context to a given task that a small, cheap model can handle it without the need for spare reasoning capacity to fill gaps.

Frontier models are expensive and companies can burn through a year’s budget worth of tokens in months. But most tasks can be handled by much smaller, more constrained models at a fraction of the cost.

“You don’t always have to pay the frontier guys for every task,” he said. “We’ve observed with some clients that we could save them 95% with a 3 billion parameter model.”

Harness

An AI agent harness is the software infrastructure scaffolding around an LLM that differentiates an agent from a chatbot. It’s what enables an agent to act on tasks rather than simply respond to prompts. A model reasons through a problem and decides what to do about it. The harness connects the model to the tools, systems, memory, guardrails, and execution environments required to perform the actions determined by the model. The term is frequently used more or less interchangeably with orchestration layer.

“It’s the process around the model, which matters way more than the model itself,” Thakrar said.

In benchmark tests, a superior harness on a less powerful model produces better results than an inferior harness on a much bigger model.

For the best results, Thakrar said, it’s essential to understand the deterministic and non-deterministic elements of a given workload, and build that into the architecture. Machines can read, organize, and validate, and they excel at deterministic tasks. Humans, on the other hand, are exceptional at non-deterministic tasks like judging, synthesizing, and deciding.

Those non-deterministic tasks in a process are the ideal point for AI agents to incorporate a human in the loop, what Thakrar calls human harness. He pointed to a stakeholder mapping agent Zoho built for sales as an example, which takes the context of an initial meeting and third-party enriched data like a LinkedIn profile, weighs probabilities, and makes an educated guess about the stakeholder map.

“The initial goal was just to eliminate that task completely from the human workflow,” he said. “The stakeholder map is done, it’s in the folder, and you can look at it.”

But the agent would struggle to capture nuance. The meanings of titles in organizations always vary, and the politics and dynamics of any given meeting can be difficult for an AI agent to discern. Rather than keep feeding the agent data to try to make it intelligent enough to make those determinations, it was simpler and more efficient for the agent to create a proposed stakeholder map and hand it over to a human who could make changes and explain why those changes were necessary.

Ultimately, Thakrar said the agent still saved human team members time because the stakeholder map was usually pretty close, and the corrections also helped the model grow smarter by adding richer context.

Specialization

Specialization is transitioning a process from testing on a frontier model to production on a much narrower, smaller model. Once you’ve proven that an agent can do a job well, you want to stop paying master-craftsman rates to keep doing that one job well.

Specialization is all about capturing your subject matter experts’ best judgement and pattern recognition to build an open-weight, open source, trained, and fine-tuned model that can be deployed in your own data center.

“The true enterprise bet is to keep that orchestration layer, which is your IP and knowledge, in house,” Thakrar said. “You don’t want to host that on someone else’s model. The goal of everyone in enterprise should be to run, train, and host their own models.”

Over the past few years, I have learned that basic prompts produce inconsistent, hallucination-prone results that no executive would trust in production. What turned the tide was my move to advanced prompting techniques. These weren’t theoretical experiments; they became a practical foundation for reliable, measurable outcomes. I want to share the techniques that consistently delivered the biggest gains in my projects, complete with real before-and-after examples, copy-paste templates, lessons from failures and guidance on when to evolve beyond prompting to agentic systems.

Why advanced prompting still matters in enterprise settings

Sophisticated prompting remains essential for control, reliability and compliance. If you “ask nicely” and hope for the best, you need deterministic behavior, auditable reasoning and minimal risk of hallucination. Here’s what worked for me.

1. Chain-of-Thought (CoT) and its variants: Unlocking step-by-step reasoning

The problem: Models would jump to conclusions on complex analysis tasks, especially involving data interpretation or multi-step logic.

What I did: I started explicitly instructing the model to “think step by step” and show its reasoning.

Before (basic prompt): “Analyze last quarter’s sales data and recommend three actions.”

After (CoT prompt):

“You’re a senior business analyst. Analyze the following sales data step by step: [data]. First, identify the key trends. Second, calculate the rates and anomalies. Third, link findings to business context. Finally, recommend the three prioritized actions with expected impact. Explain your reasoning at each step.”  

Results: Accuracy and depth improved dramatically.

Variants that worked well: Self-consistency. I ran the same CoT prompt multiple times and took the majority consensus. This reduced variability significantly.

Template you can use:

You are [expert role]. Solve this problem by thinking step by step.

[Task or question]

For each step:

1. State your observation or calculation.

2. Explain the implication.

3. Proceed only when confident.

Final answer in this format: [structured output]

2. Tree-of-Thoughts (ToT): Exploring multiple reasoning paths

For truly complex decisions such as resource allocation or risk assessment, linear CoT isn’t enough. Tree-of-Thoughts lets the model generate and evaluate multiple branches.

Example: I was helping a client evaluate three potential vendor platforms for an AI deployment. A standard prompt gave a superficial comparison. With ToT

Prompt Snippet:

Explore three different reasoning paths for selecting the best vendor platform:

Path 1: Focus on cost and scalability.

Path 2: Focus on security, compliance and integration.

Path 3: Focus on innovation and long-term roadmap.

For each path, evaluate pros/cons against our requirements [list].

Then, compare the paths and recommend the strongest overall option with justification.

Outcome: The model surfaced nuanced trade-offs (e.g., one vendor had superior security, but higher integration cost).

When to use: Strategic planning, troubleshooting or scenarios with high uncertainty and multiple viable approaches.

3. ReAct (Reason+ Act) and prompt chaining: Moving toward agentic behavior

One of the biggest leaps I have noticed comes from combining reasoning with tool use and chaining prompts.

ReAct example: (used in data analytics workflow)

You are an AI analyst with access to tools. For the query below:

1. Reason about what information you need.

2. Choose the appropriate tool or action.

3. Observe the result.

4. Repeat until you can answer confidently.

Query: [user request]

In practice, I chained this with retrieval tools. One automated quarterly compliance reporting; the system reasoned about required data, pulled relevant records, validated them, and generated the reports.

4. Meta-prompting and self-reflection: Letting the model improve itself

Use the model to refine its own prompt. This is a huge time-saver.

You are an expert prompt engineer. Improve the following prompt for clarity, structure and effectiveness with [target model]. Make it more precise while preserving intent.

Original prompt: [paste]

Provide the improved version and explain your changes.

Self-reflection loops (asking the model to critique its own output and revise) are a game-changer for content generation and code-review tasks.

5. Multimodal and structured output techniques

With vision-enabled models, I started combining text with images (e.g., uploading architecture diagrams or dashboards).

Tip from experience: Be extremely specific in describing what the models should focus on.

Best practices I learned the hard way

• Start simple, then layer complexity: Over-engineered prompts from Day One usually backfire.
• Model specific tuning: Some models respond better to XML delimiters; others to explicit reasoning.
• Evaluation and versioning: Treat prompts like code if you track versions and run automated evals.
• Security guardrails: Always include instructions against prompt injections and respect data boundaries.
• When to stop prompting: For repetitive, high-stakes workflows, move to full agents or an orchestration framework.

Final takeaways for technical leaders

Advanced prompt engineering has now become a core competency for anyone responsible for enterprise AI outcomes. Start by picking one technique and apply it rigorously to a real business problem. Document before/ after and you will notice why it’s worth mastering.

The field continues evolving towards more automated and agentic systems, but the ability to precisely direct AI reasoning remains foundational.

This article is published as part of the Foundry Expert Contributor Network.
Want to join?

Cloudflare recently beat Q1 2026 earnings. Revenue up 34% year over year. EPS ahead of consensus. Full-year guidance raised. Then, in the same breath, they announced 1,100 layoffs, 20% of the company. CEO Matthew Prince’s explanation: “The way we work at Cloudflare has fundamentally changed.”

Block did the same thing. Beat guidance, raised outlook, cut 4,000+ jobs. Both framed it as architecting for the AI era.

This is not a contradiction. This is the new math boards are running. And if you’re a CIO who hasn’t started running it yourself, you’re behind.

The benchmark has moved

AI-native companies have quietly reset what “efficient” means for a technology organization. Midjourney generates over $500M in revenue with roughly 160 employees, over $3M per head. Anthropic hit a $14B annualized run rate in early 2026 with fewer than 3,000 employees. Across the top AI-native startups, the average revenue per employee is $3.48M, nearly twelve times the traditional SaaS benchmark of $300K.

Boards aren’t comparing you to your 2019 self anymore. They’re comparing you to Anthropic.

This is the pressure Cloudflare and Block are responding to. They’re not cutting people because the business is struggling. They’re cutting because investors have internalized a new denominator. Headcount is no longer a proxy for capacity; it’s a liability on the efficiency ratio.

For CIOs, this creates a hiring problem that looks nothing like the cloud or mobile talent gaps of the past decade. Those gaps were about volume: hire 100 cloud engineers, absorb the cost, build the capability… This one is about density; you’re not looking for 100 people. You’re looking for 10 who can deliver what 100 couldn’t, and justify $1M or more in value per seat.

Finding bodies to fill seats has never been easier. Finding people who operate at that level of leverage is a different problem entirely.

‘Acqui-hires’ are a shortcut with a hidden cost

Companies have figured out that recruiting AI-native talent one by one is too slow and that it’s faster to buy a team. Google’s acquisition of the Windsurf founders, Meta bringing in the Scale AI team, Accenture’s string of AI-focused acquisitions: these are acqui-hires dressed up as M&A. The premium on experienced AI talent is high enough, and the urgency real enough, that organizations are skipping traditional hiring loops entirely and buying their way in.

I’ve been on the other side of this. My company, MadKudu, was acquired by HG Insights specifically to bring AI-native capability into an established enterprise business. HG needed change agents who had already figured out how to build and ship in this new era, not just people who’d read about it. That’s the thesis behind most of these deals.

But there’s a cost that doesn’t show up in the acquisition price.

AI-native teams are fast because they operate with a different set of defaults: full access to tools, minimal governance layers, the ability to experiment and ship without a six-week approval cycle. That operating model is not a perk; it’s the fundamental mechanism. It’s why a team of 10 can do what an enterprise team of 100 can’t.

When you acqui-hire that team and then slot them into your existing approval processes, you’ve bought the people and killed the engine. The change agents you paid for become change-frustrated. The attrition that follows is expensive and predictable.

The harder realization: acquiring an AI-native team means accepting how they work. That requires deliberately carving out space for them to operate differently, not just tolerating it but institutionalizing it. The acquisition is an organizational change program, not just a hiring event.

The CIO’s real problem

The governance stack most enterprise organizations run was designed for a headcount world. Every tool vetting cycle, every vendor review, every security approval was calibrated assuming you were managing a large team where consistency and control were the primary objectives.

That calculus breaks when your goal is talent density. The same approval processes that protect against data leaks are now the reason your best people can’t do their best work. When it takes six weeks to approve a tool that your competitor’s team is already shipping with, you’ve traded velocity for the perception of safety.

The practical fix is structured experimentation: clear guardrails, defined boundaries, but explicit permission to try tools before deciding whether to roll them out broadly. Gating everything prevents you from ever discovering what 10x productivity looks like.

The skills inventory question is also more nuanced than it sounds. Job titles won’t tell you where the leverage is. You need to map the actual tasks within each function and assess which can be automated or augmented with AI. That’s where you find the people who, with the right tools, become your $1M/employee talent, not because you hired differently, but because you enabled better.

This is also where the build-versus-buy question gets genuinely tricky. As AI reshapes how products are built and delivered, your internal operating model — how you work, how fast you ship, how you use data — is becoming core IP. Outsourcing delivery means outsourcing the part of the organization where your competitive advantage is now being built.

Closing the gap without slowing down

The AI talent wars are not primarily a recruiting problem. They’re a rethinking of what organizations are supposed to look like.

Boards have a new benchmark. Cloudflare, Block, Amazon, Meta and others have already started restructuring to meet it, publicly, painfully, even while beating their numbers. The question for CIOs isn’t whether this pressure arrives; it’s whether you’re ahead of it or behind it when it does.

The organizations that navigate this well won’t win by outbidding competitors for a handful of elite engineers. They’ll win by designing operating systems that amplify the leverage of the talent they do have, by enabling their best people rather than constraining them, and by treating AI fluency as a core organizational capability rather than a niche specialization.

Talent density is the new headcount model. The sooner your governance, your tooling and your board conversations reflect that, the better positioned you’ll be when the next efficiency report lands.

This article is published as part of the Foundry Expert Contributor Network.
Want to join?

Four months of Claude Code — that’s all it took for Uber to burn through its entire annual budget for AI. Token after token, engineers embraced the platform with few control mechanisms tying costs to outcomes. The result was a budget runaway and a clear case study in how limited oversight snowballs into an AI blowout.

This is a phenomenon I like to call “shadow tokens” — AI credits paid for by the company but largely invisible to decision-makers. Too many engineers have the final say over how much they consume and, therefore, what it costs. This all-you-can-eat attitude is part of the reason why Microsoft is reportedly winding down many internal licenses across key engineering teams and why one in five organizations is missing its AI spend forecast by more than 50%.

And the trend is only accelerating. By 2028, Gartner predicts that AI coding costs (driven by this kind of ungoverned consumption) will be as much per developer as the salary companies pay that person.

LLMs and agents introduce a new class of variable cost that scales with behavior rather than headcount, putting enterprises on the hook for tools that balloon with workload. I don’t see this as enterprises overspending because they’re reckless — it’s down to a lack of managerial oversight, budget alignment that demands a proven return on investment, and engineer education on how much is too much.

Going forward, CIOs need to thread the AI needle between governance that encourages transparency and reasonable spend without stifling innovation.

When shadow tokens result in real costs

The issue is that AI isn’t a traditional line item. Previously, enterprise leaders onboarded software-as-a-service (SaaS) with a good idea of the total cost. An allocated software seat or annual contract was a known quantity. The cloud added some variation (with fluctuations depending on hosting size), but instances were still modelable. AI flips this status quo on its head — the unit of consumption is behavior and the cost is exponential.

And these specifics aren’t immediately apparent at pilot. Tools can appear inexpensive in controlled experiments yet unpredictably scale depending on session length, context window size, model selection and whether agents run in parallel. This is the fallacy of the $20-per-seat enterprise plan — tokens are charged separately at API rates with no ceiling. The final dollar value of any session is set by factors that finance can’t always model in advance, particularly when these decisions usually rest with the engineers themselves.

According to Deloitte, only 21% of organizations deploying agents have a mature governance model, a real concern because they’re token-eating machines. This is what was happening at Uber — Claude Code in agentic mode was autonomously reading codebases, planning changes across dozens of files and opening pull requests. Each step quickly adds up, with Anthropic’s own documentation noting that agents consume approximately seven times as many tokens as standard sessions.

This is shadow IT and shadow AI, evolved. This time, however, many leaders approved the tool in question without guardrails governing consumption. AI hype adds fuel to the fire and normalizes long sessions. Uber’s CTO, for example, described a company-wide shift toward “agentic software engineering” with employees “who are quietly experimenting, quietly shipping and quietly pushing things forward”. This is an exciting way to test the limits of what’s possible, certainly, but it’s also a position that goes a long way to explaining how the company spent its annual AI budget by April.

Shifting the culture from usage to yield

Engineers haven’t done anything wrong here. In fact, they’re adopting and experimenting as instructed, with Uber creating leaderboards and ranking users by token consumption. More use led to a better ranking, reflecting a culture that lauds new ways of doing things. This behavior is known as “tokenmaxxing,” and its principal knock-on effect is shadow tokens — quantity-over-quality processes that leaders struggle to control until they’re fully realized in the budget. Of course, if management treats adoption metrics as performance metrics, then engineers can’t be blamed for using more tokens. The tension is that the teams driving adoption aren’t the ones managing spend.

None of this is meant to dismiss AI’s productivity possibilities and potential return on investment. Developers save 3.6 hours per week, achieve 60% higher pull request throughput and cut onboarding time in half with automation. Meanwhile, Uber shared that roughly 11% of live backend updates were written by agents with no human in the loop. However, these wins aren’t the problem — it’s that too many teams aren’t connecting input to output. I’ve spoken to admins who discovered their token spend had tripled in a single quarter after using heavier models or accidentally doubling up on agentic applications. Nobody knew until the financial damage was done.

Automation needs to happen sustainably with an eye on the bottom line. In my view, a much better metric for achieving this is AI yield — the measurable business or engineering output generated per dollar spent on tokens. Otherwise, without a feedback loop, even genuinely productive teams are flying blind.

Stopping token waste before an AI blowout

Creating that throughline between AI investment and token consumption starts with established financial metrics. This is possible via maximum spend limits (dictated by spend tagging, workload tiering and cost-per-output benchmarks) per team or project. Then, any additional allocation requires approval, closing the loop between the engineers spending the tokens and the leaders paying for them. AI isn’t cheap and teams should demonstrate a bang for their buck.

This is something we do with our engineering team at Hexnode. Resource allocation for Claude Code and Cursor is tied directly to ROI rather than letting consumption run open-ended. Given the pay-as-you-go nature of these tools, a firm usage limit per team offers simple but essential control.

Similarly, there’s room to apply some of the governance principles IT uses for device management. Things like policy enforcement, role-based access, real-time monitoring and automated alerts can flag usage behavior in advance. Uncovering such insights at the token layer works to identify power users and prevent excessive spending.

We also need to encourage cultures that praise outputs that actually achieve efficiency. AI applications that result in shipping faster, reducing rework and cutting review cycles are gains that should be celebrated. If your company hosts leaderboards, frame unnecessary token burn as wasteful rather than valuable. The organizations creating healthier consumption habits work with their engineers to understand not just how to use AI, but what responsible use looks like and what it costs.

This is a conversation teams need to have now. Anthropic just ended flat-rate pricing for programmatic workloads from June 15. Now, agents, continuous integration pipelines and automated workflows draw from a dedicated monthly credit pool billed separately from the subscription. Once that pool is exhausted, agent tasks either stop entirely or overflow to extra billing. Work can either get very expensive or grind to a halt for teams that aren’t prepared.

Getting a grip on shadow tokens means better rules and tools connecting spend to outcomes. Only by building the financial and cultural infrastructure that encourages sustainable adoption can leaders see what they’re spending, connect it to what they’re getting and course-correct before the costs become a crisis. Ultimately, shadow tokens are only invisible if we choose not to look.

This article is published as part of the Foundry Expert Contributor Network.
Want to join?

In August 2025, 126 malicious packages landed in the npm registry. Even after the community caught the initial wave, 80 of these hidden backdoors remained actively listed.

That was enough. Over 86,000 downloads. Malicious code in PhantomRaven, packages running in the production systems of Fortune 500 companies worldwide. And throughout the entire window, not a single EDR/XDR alert.

This happened because the attack surface has expanded to a layer EDR/XDR was never designed to see: VS Code extensions, local MCP servers, and rogue AI coding assistants that inherit your engineers’ valid credentials to steal data at machine speed.

To eliminate this structural vulnerability, Palo Alto Networks acquired Koi, an AI-native developer security product engineered for proactive, precision enforcement. Below we compiled a 2026 CISO checklist you can use to audit your environment and see how Koi automates each defense from day one.

#1. Gain real-time visibility into shadow AI & extensions

Your existing asset management tracks binaries and installers, but it cannot see local VS Code extensions, MCP servers, or ad-hoc Python scripts running on developer endpoints. This visibility gap was recently exposed by the MaliciousCorgi campaign, where two marketplace extensions with 1.5 million combined installs silently harvested every file a developer opened. Neither triggered any detection because they were not binaries, not executables, not anything your inventory was built to flag. To counter this, Koi closes the gap by analyzing what extensions actually do after installation, exposing hidden data-harvesting channels running inside your active workspace.

#2. Distinguish between human and autonomous agent behavior 

When a rogue AI agent exfiltrates your proprietary source code, it uses a developer’s valid credentials during normal working hours, making the session look entirely legitimate to standard XDR baselines. Moving beyond static permission lists, Koi deploys behavioral profiling within the workspace runtime. By actively intercepting unauthenticated background tasks and blocking unauthorized file-system reads, it stops automated data exfiltration in real time.

#3. Establish guardrails for automated package updates on endpoints

Developers prioritize speed, often allowing software packages to auto-update on their endpoints the moment a new version appears. Attackers weaponize this supply chain vulnerability, as seen in the May 2026 Team PCP attack where 3,800 GitHub repositories were compromised in just 36 minutes via poisoned auto-updates. Securing agentic endpoints against these rapid breaches requires behavior-based inspection within the active workspace context. Koi operates at this layer by providing safe deployment buffers that automate version cooldowns, blocking bleeding-edge updates until they are vetted. By continuously auditing process creation within the IDE runtime, Koi instantly drops unauthorized remote connections before malicious payloads can exfiltrate credentials from the endpoint.  

#4. Enforce principle of least privilege for AI agents

AI coding assistants inherit the privileges of whoever deployed them. In practice, that means read access to production databases, write access to core repositories, and access to every secret in environment files and configuration directories. To restrict this excessive access, Koi applies dynamic sandboxing directly to AI agent processes at the kernel level. It enforces a strict zero-trust boundary that segregates sensitive workspace vectors, preventing agents from pulling data outside their approved scope without interrupting developer workflows.

#5. Maintain continuous endpoint posture management

Signature-based scanning only stops known threats. Sophisticated repository attacks often arrive as functional, high-rated software that carries no known bad signature. Koi’s research into the DarkSpectre campaign found eight browser extensions, all carrying “featured” badges from Google and Microsoft, installed by over 8 million users, silently harvesting every conversation from ChatGPT, Claude, and Gemini in the background. Koi addresses this by operating upstream: scanning marketplace listings every hour, using LLM-driven code analysis to compare what software promises against what its code does, sandboxing it, and scoring the risk before it ever reaches the endpoint.

Summary

Securing the modern enterprise is no longer about patching individual gaps. As AI agents redefine the workforce, Agentic Endpoint Security (AES) is now a strategic imperative for every CISO. By establishing a mandatory control plane for the AI-native workspace, AES ensures that your organization can scale engineering velocity without ever compromising enterprise integrity. 

Ready to secure the future of your software stack? See how Koi Agentic Endpoint Security delivers complete visibility, risk scoring, and real-time prevention across every endpoint in your enterprise.