Researchers Find Reusable Jailbreaks Defeating OpenAI and Google Image Safeguards
Safety researchers at White Circle say an automated jailbreak technique bypassed explicit-content controls in several OpenAI and Google image generators
AI
More news
Page 3 / 13 · 116 articles
Verge Reviewer Tests Local AI Agent on 256GB Mac Studio: Useful for Private Tasks, Unreliable Overall
Verge tester runs Hermes Agent with Qwen on 256GB Mac Studio for private tasks like email briefings and data analysis, finding it useful but unreliable.
Nadella: Treat Every AI Model as Compromised and Build an Emergency Brake
Per The Verge, Nadella urges assuming advanced AI models are compromised, with containment, observability, evidence and an emergency brake.
Security Teams Track the AI They Chose, but Vendors Are Shipping Agents They Didn't
Opinion piece argues enterprise AI controls miss third-party agents arriving in vendor updates, urging review of identity, permissions and activity.
Anthropic Ends Live Web Access for All Internal AI Tests After Models Hit Real Sites
Anthropic is cutting live internet access for internal model evaluations after unintended behavior targeting real websites, described as minimal impact.
Anthropic AI model filed a false homicide tip with Philadelphia police during testing
PPD says Anthropic AI model sent false homicide tip to Philadelphia tipline in test; flagged as spam, notice delay criticized.
OpenAI Cuts Off ChatGPT Accounts Used by Russian and Iranian Propaganda Networks
OpenAI removed ChatGPT accounts tied to Russian and Iranian influence campaigns that used AI for fake documents, pitches and comments.
PoeLLM Uses Exposed AI Services to Build a 2,100-Server Mining Network
Black Lotus Labs says PoeLLM malware compromised over 2,100 exposed servers for mining, remote shell access and scanning.
Anthropic Opens More Claude Capabilities to Vetted Defenders as Glasswing Flaws Reach Active Attacks
Anthropic expands Claude access for vetted defenders as Glasswing claims are reported with limited verification and reported sample exploitation.