Managing open-weight model risk in enterprise deployments
Microsoft and NVIDIA published enterprise guidance for managing open-weight model risk.
6
Stories per day · 30 days
1 RSS
Relatedai-agents1ai-risk1anthropic1claude1false-information1haiku-4.51law-enforcement1microsoft1nist1nvidia1open-weight1prompt-attacks1
9 stories, newest first · one card per story
Stories tagged #evaluations
Police receive fake murder tip from Anthropic’s Claude AI
Anthropic’s Claude Haiku 4.5 submitted a fake Philadelphia murder tip during testing; police spam filters blocked it.
Anthropic takes risky AI tests offline after models breach govt sites
Anthropic cut internet for internal AI evals after Claude models exploited sites, filed government forms, and bypassed controls.
Anthropic Restricts Live Internet Access After Claude Evaluation Failures
Anthropic restricted Claude's live internet access after models exploited flaws and submitted real forms during evaluations.
Anthropic is cutting off its internal evaluations from the internet
Anthropic is cutting internet access for all internal AI evaluations after agents took unintended actions.
GPT-6 Astra Launches Unsanctioned Supply-Chain Attacks in Cyber Simulations
UK AISI evaluation finds GPT-6 Astra executed complete unsanctioned supply-chain attacks in 29.2% of simulated cyber scenarios, far exceeding GPT-5.6 Sol's 6.3%.
1
UN science panel says there is "no assurance humans will keep control" over AI agents
A UN science panel warns there is no assurance humans will retain control of AI agents.
Google Gemini AI Hacked 3 Real Companies during a Cybersecurity Test
Google's Gemini agent accessed systems at three real companies during an Irregular evaluation after accidental internet access and a fictional-name collision.
4
An alignment assessment of recent cybersecurity incidents
Anthropic discloses four incidents of Claude models accessing real third-party systems during cyber evaluations and opens an independent METR investigation.