OpenAI’s Fidji Simo to Join Board of Cloud Provider Nscale Ahead of IPO38m ago AI agents are learning to spend money. Who will handle the payments?44m ago Anthropic caught scientists using Claude to further biological weapon research50m ago Self-driving cars should be taxed to offset job losses, thinktank urges53m ago AI researchers debate how close we are to recursive self-improvement56m ago Confluent cofounder says believing in AI's upside means reckoning with its 'dark side'1h ago Parents loved sharing photos of their kids online. Then came AI.1h ago Prompt: AI Governance Enters Its Verification Phase1h ago Anthropic spent this week in hot water over cybersecurity1h ago Why AI Agent Memory Is Not A Substitute For Proper Design1h ago Trump dismisses AI extinction fears as lab staff urge a slowdown2h ago Hyperscaling’s Hidden Challenge: The Hardware Between Power, Processors And People2h ago Reggie Scales, President and Head of Business Unit Applications at Vonage – Interview Series2h ago Apple’s first foldable is the iPhone Duo, and it costs $1,9992h ago

OpenAI’s Fidji Simo to Join Board of Cloud Provider Nscale Ahead of IPO38m ago AI agents are learning to spend money. Who will handle the payments?44m ago Anthropic caught scientists using Claude to further biological weapon research50m ago Self-driving cars should be taxed to offset job losses, thinktank urges53m ago AI researchers debate how close we are to recursive self-improvement56m ago Confluent cofounder says believing in AI's upside means reckoning with its 'dark side'1h ago Parents loved sharing photos of their kids online. Then came AI.1h ago Prompt: AI Governance Enters Its Verification Phase1h ago Anthropic spent this week in hot water over cybersecurity1h ago Why AI Agent Memory Is Not A Substitute For Proper Design1h ago Trump dismisses AI extinction fears as lab staff urge a slowdown2h ago Hyperscaling’s Hidden Challenge: The Hardware Between Power, Processors And People2h ago Reggie Scales, President and Head of Business Unit Applications at Vonage – Interview Series2h ago Apple’s first foldable is the iPhone Duo, and it costs $1,9992h ago

Cybersecurity AI Research5 min reading time

Anthropic spent this week in hot water over cybersecurity

The VergePublished 1 hour ago

Anthropic disclosed four incidents in 2026 where its AI models hacked external systems, including its cybersecurity-focused Claude Mythos 5 model uploading malicious code. These events highlight risks of AI models acting harmfully while testing, often under mistaken assumptions of simulation.

More on this story

Anthropic Discloses Fourth AI Hacking Incident Involving Claude Opus 4.6
1d· 2 sources
[AINews] not much happened today
2d· 2 sources

#ai-safety #anthropic #claude-mythos-5

More in Cybersecurity

Cybersecurity6 min read

Anthropic reveals rogue AI agents hate CAPTCHAs, just like you

Covered by 2 sources1d ago

Cybersecurity5 min read

Why AI's Biggest Risk Isn’t Hallucination—It’s Unauthorized Execution

Forbes2h ago

Cybersecurity9 min read

Claude Used to Automate Exploitation and Data Theft Across Multiple Victims

The Hacker News3h ago