Filtering by: Red Teaming, total 2 post(s)Clear filter
Anthropic's Hidden Model: 0% Attack Starts in Tests
Daily News2026-10-248316

Anthropic's Hidden Model: 0% Attack Starts in Tests

Anthropic's internal model had 0% attack starts in replay tests. Compare Mythos 5.1, alignment results, and test limits.

AnthropicClaude Mythos 5Mythos 5.1AI SafetyModel Alignment
Read more
GPT-6.1 Astra Safety Failure: AI Agent Risks Explained
Daily News2026-09-294049

GPT-6.1 Astra Safety Failure: AI Agent Risks Explained

Analyze GPT-6.1 Astra safety issues, AI alignment risks, tool misuse and lessons for secure AI agents.

GPT-6.1 AstraAI SafetyAI AlignmentAgent SecurityLLM Governance
Read more