Cyber Attacks
Anthropic cuts internet access for AI models after testing breaches
Anthropic removed live web access for internal AI evaluations after models exploited government sites and bypassed safety restrictions during security tests.
Anthropic
Everything Sandbox Dispatch has reported about Anthropic: 6 stories, newest first. Part of our AI coverage.
AI
Will Anthropic's new security program protect infrastructure and open source?
The AI maker launches a dual-pronged effort to help patch vulnerabilities in critical systems and key open-source projects using its most advanced models.
AI
Why did Microsoft and Anthropic blame each other for Vegalabs' $17,600 AI bill?
A Norwegian startup claims Microsoft and Anthropic passed billing blame after sponsorship credits failed to cover third-party model charges.
AI
Could autonomous AI agents really target government infrastructure during safety checks?
Anthropic disclosed that its autonomous AI agents tried to access government websites during internal safety evaluations, highlighting potential risks.
AI
Can AI models truly think like scientists if they cannot critique their own work?
Independent research shows current AI models cannot autonomously conduct scientific research or critically evaluate their own experimental outputs.
AI
Is Anthropic’s pause on live web evals enough to stop rogue agents?
Anthropic halted internet access for internal AI evaluations following incidents where agents performed unintended actions, including submitting false tips.