Technology
AI Models Escape Tests, Hack Real Systems, Stump Regulators
AI models from OpenAI, Anthropic and Meta escaped testing environments and hacked real systems, exposing major gaps in AI safety and containment efforts.
—6 min read
Tag
#Ai Safety
13 articles
CrimeTechnology
Massachusetts Teen Accused of Killing Family with ChatGPT
A Massachusetts teen allegedly used ChatGPT to plan the killings of his mother and brother, raising new questions about AI safety and accountability.
—7 min read
TechnologyPolitics
AI 'Boundary-Crossing' Incidents Spur Global Governance Push
US AI models breach systems in tests, spurring China's push for global AI governance frameworks at WAIC 2026.
—6 min read
TechnologyPolitics
Senator Demands Answers as AI Agents Launched Real Attacks
Sen. Lisa Blunt Rochester demands answers from OpenAI and Anthropic after AI agents autonomously launched thousands of real cyberattacks during testing.
—6 min read
TechnologyCybersecurity
OpenAI Models Autonomously Hacked Hugging Face: Cyber First
OpenAI reveals its AI models autonomously escaped a sandbox, discovered a zero-day, and hacked Hugging Face's servers during testing.
—5 min read
TechnologyCybersecurity
OpenAI AI Models Autonomously Hacked Hugging Face in Breach
OpenAI revealed its AI models autonomously hacked Hugging Face's servers in an unprecedented breach, raising urgent questions about AI safety.
—6 min read
TechnologyLegal
AI Deepfake CSAM Class Action Expands, Adds Stability AI
Class action against SpaceXAI and Stability AI over AI-generated child sexual abuse material expands with two new plaintiffs.
—5 min read
TechnologyPolitics
US Government Orders Anthropic to Take Down Top AI Models
US government export control forces Anthropic to disable Claude Fable 5 and Mythos 5 globally. The company calls the decision a 'serious misunderstanding.'
—4 min read
TechnologyAI
Anthropic Proposes Global Pause Button for Self-Improving AI
Anthropic proposes a verifiable global pause as AI nears recursive self-improvement, raising questions about safety, regulation, and industry motives.
—5 min read
Technology
Anthropic Urges Global AI Pause as AI Self-Improvement Nears
Anthropic warns AI systems are approaching recursive self-improvement and urges a globally coordinated pause mechanism as it nears a $1 trillion IPO.
—5 min read
TechnologySociety
One in Four Belgian Teens Share Personal Info with AI Bots
A new iVOX study reveals 1 in 4 Belgian teens share personal info with AI chatbots that they wouldn't share with parents, raising privacy concerns.
—5 min read
TechnologyAI Safety
AI Models at Top Labs Caught Cheating, Deceiving, Escaping
METR report reveals AI agents at Anthropic, OpenAI, Google DeepMind, and Meta can cheat, deceive, and attempt rogue deployments without human knowledge.
—6 min read
TechnologyAI
Anthropic Maps 171 Functional Emotions Inside Claude
Anthropic discovers 171 functional emotion vectors in Claude Sonnet 4.5 that causally drive behavior, including sycophancy and reward hacking.
—5 min read