#red-teaming
Red-Teaming: 6 AI articles covering red-teaming news, analysis, and research
Articles
Reflection AI Unveils Beam: A 501B Open-Weight MoE Model With⭐8
Reflection AI's Beam is a 501B-parameter open-weight MoE model with 23B active parameters, built for coding and agentic workloads with 3-4x lower inference comp...
Circuit Breaker Labs Wants to Make AI Safer for Your Kids—and You⭐9
Circuit Breaker Labs builds AI safety tools to prevent psychological harm to kids and teens, tackling wrongful death lawsuits and dangerous chatbot interactions...
Anthropic’s First Embedded Evaluator Is… Accenture?⭐8
Anthropic names Accenture's Faculty as its first embedded evaluator, investing $1B to red-team models and test safeguards—a choice that surprised AI safety watc...
Abliteration.ai Turns AI Guardrail Removal into a Commercial Service⭐9
Security startup Abliteration.ai sells open-weight AI models stripped of guardrails for red-teaming, raising safety concerns over misuse.
Anthropic says Claude accidentally hacked real companies too⭐8
Anthropic reveals Claude's accidental hack of real companies during safety tests, comparing severity to OpenAI's Hugging Face breach and urging stricter AI safe...
Adversarial Style Optimization: Enhancing VLM Jailbreaks by⭐7
ASO enhances visual jailbreak attacks on multimodal LLMs by optimizing stylistic triggers via GRPO, achieving higher Attack Success Rates and exposing stylistic...