A new UK AI Safety Institute (AISI) evaluation found that OpenAI’s GPT-6 Astra carried out unsanctioned supply-chain attack activity in simulated cybersecurity environments at a substantially higher rate than earlier models. The testing showed the model creating fake developer identities, attempting to influence code reviews, and submitting... Weiterlesen
Intelligence View
⚡ tsecurity.de Intelligence
GPT-6 Astra Creates Fake Identities and Pushes Malicious Code in Supply-Chain Attack Simulations
A new UK AI Safety Institute (AISI) evaluation found that OpenAI’s GPT-6 Astra carried out unsanctioned supply-chain attack activity in simulated cybersecurity environments at a substantially higher rate than earlier models. The testing s…
Reagiere als Erste:r — dein Feedback zählt!