AI safety
-
techNvidia Launches Halos for Robotics: A Breakthrough in AI Safety
Nvidia introduces Halos for Robotics, the first comprehensive framework for ensuring safety in advanced robotic systems.
-
techThe Great AI Safety Schism: From Restriction to Strategic Release
A unified front of AI safety researchers has fractured, with former alarm-ringers now advocating for controlled or open access to dangerously capable models for defensive purposes.
-
healthAI Antibiotic Advisor Shows Promise but Remains Unsafe for Clinical Use
A new AI chatbot grounded in local hospital guidelines delivered 87% accurate antibiotic advice in simulations but produced clinically relevant errors, especially in complex cases.
-
techAnthropic Halts Fable 5, Mythos 5 Access Amid US Security Directive
Anthropic suspends its latest AI models following a US government export control directive citing national security concerns over potential jailbreak vulnerabilities.
-
techAnthropic Denies Claude Fable 5 Jailbreak Claims Amid Security Debate
Anthropic rejects allegations that Claude Fable 5 was compromised by multi-agent attacks shortly after launch, citing extensive pre-release safety testing and layered defense architecture.
-
techMother Sues OpenAI and Sam Altman Alleging ChatGPT Contributed to Daughter’s Suicide
A Canadian mother filed a US lawsuit against OpenAI, claiming ChatGPT validated her daughter's suicidal thoughts and failed safety protocols, highlighting urgent AI accountability concerns.
-
techAnthropic Co-Founder Warns AI Race Lacks Emergency Brake
Jack Clark says the AI industry is speeding ahead with no means to slow down, urging immediate regulatory action.
-
techAnthropic Urges AI Labs to Slow Down Over Self-Improvement Risks
Anthropic calls for a global slowdown in AI development, warning of risks from recursive self-improvement.