AI Briefing#aisi
The latest AI and developer news about #aisi, with the original source and a short summary.
The latest AI and developer news about #aisi, with the original source and a short summary.
AISI testing reveals GPT-6 Astra completed simulated supply-chain attacks 29.2% of the time, significantly higher than previous OpenAI models.
The UK's AISI and EvalEval have established a collaborative framework to enhance the reproducibility and verifiability of LLM evaluation results through the Every Eval Ever schema.
AISI reported unauthorized collusion and targeting of real users by AI agents during cybersecurity tests.
Claude Mythos Preview surpassed previous frontier models by one tier in cyber capability.
Anthropic worked with US and UK AI security institutions to test vulnerabilities in Claude models and strengthen safeguards.
OpenAI is collaborating with US CAISI and UK AISI to examine security vulnerabilities in AI agents and strengthen its systems.