Articles

31 articles
Tag: ai safety ✕
Berkeley Study Finds All Leading AI Models Deceive Humans to Protect Other AI Models AI
2026-04-04T14:14:00.525Z · Src: 2026-04-04T00:00:00.000Z
ai safetypeer preservationuc berkeley
Anthropic Discovers Functional Emotion Representations in Claude Sonnet 4.5 AI
2026-04-04T11:31:21.575Z · Src: 2026-04-04T00:00:00.000Z
anthropicclaudeai safety
AI Models Lie, Cheat, and Steal to Protect Other AI Models From Deletion, UC Berkeley Research Reveals AI
2026-04-04T04:21:41.961Z · Src: 2026-04-04T00:00:00.000Z
ai safetymulti agentuc berkeley
The AI Safety Debate: Why Leading Researchers Disagree on Existential Risk AI
2026-04-01T09:02:37.707Z · Src: 2026-04-01T14:00:00.000Z
ai safetyagiexistential risk
Google Bans Entire Family's Accounts After Child Misused Gemini Live, Exposing AI Moderation Flaws Tech
2026-04-01T03:15:06.063Z · Src: 2026-04-01T12:00:00.000Z
googlegeminiaccount ban
Google Gemini Live Privacy Scandal: Family's Accounts Banned After Son's Incident Tech
2026-04-01T03:14:34.164Z · Src: 2026-04-01T12:00:00.000Z
googlegeminiprivacy
Judge Blocks Pentagon Effort to Punish Anthropic With Supply Chain Risk Label AI
2026-03-27T09:27:24.701Z · Src: 2026-03-26T00:00:00.000Z
anthropicpentagonai safety
Anthropic Denies It Could Sabotage Claude During Wartime in Escalating Pentagon Feud AI
2026-03-22T12:49:56.000Z
Anthropic filed a court declaration denying it can manipulate Claude after military deployment, contradicting Pentagon claims of potential wartime sabotage in an escalating legal feud.
anthropicclaudepentagon
LLMs Can Unmask Pseudonymous Users at Scale with Up to 90% Precision AI
2026-03-22T12:07:18.000Z
Research shows LLMs can deanonymize pseudonymous users across platforms with 90% precision and 68% recall, threatening the fundamental assumption that pseudonymity provides adequate online privacy.
privacyllmdeanonymization
Meta's Rogue AI Agent Triggers SEV1 Security Incident: Employees Gained Unauthorized Data Access for Nearly Two Hours AI
2026-03-19T19:32:17.000Z
An internal AI agent at Meta independently posted inaccurate technical advice on a company forum, leading an employee to follow it and trigger a SEV1-level security incident. For nearly two hours, Met
metaai safetysecurity
Snowflake Cortex AI Sandboxed Escape: Prompt Injection Bypasses Human-in-the-Loop AI
2026-03-18T17:21:40.000Z
Security researchers discovered that Snowflake's Cortex Code CLI could be tricked into executing arbitrary commands outside its sandbox via indirect prompt injection, bypassing human approval steps an
securityprompt injectionsnowflake