#red-teaming
49 related articles
·3 min
From Claude Oceanus to GPT-5.6: A Complete Breakdown of This Week's Major AI Model Updates
Deep analysis of this week's major AI model updates: Anthropic Oceanus red team leak, OpenAI GPT-5.6 Dual Alpha exposed, NVIDIA Nemotron Ultra 550B release, and AI recursive self-improvement research breakthrough.
Read more →
·2 min
Jan Leike Launches New Research Project at Anthropic: Alignment Is Only Part of AGI Safety
Former OpenAI Superalignment lead Jan Leike announces a new research project at Anthropic, stating AGI safety goes far beyond alignment alone.
Read more →
·1 min
PNAS Study: Human Persuasion Techniques Can Manipulate AI, Raising Compliance Rate from 35% to 51%
A new PNAS study finds classic human persuasion techniques can effectively manipulate LLMs, raising AI compliance with inappropriate requests from 35% to 51%, revealing human-like psychological weaknesses in AI.
Read more →
Tech Frontiers·2 min
GPT-5.6 First Leak: How Self-Training Loops Enable OpenAI's Three-Week Iteration Cycle
OpenAI's GPT-5.6 has entered internal testing, just three weeks after GPT-5.5. The key accelerator is the self-training loop introduced in GPT-5.3, enabling exponential iteration speed.
Read more →
Tech Frontiers·1 min
Anthropic Donates AI Alignment Tool Petri to Meridian Labs: A New Chapter for Open-Source Safety Evaluation
Anthropic donates AI alignment tool Petri to Meridian Labs with a major update improving adaptability, realism, and depth. Analysis of the impact on AI safety.
Read more →
Tech Frontiers·1 min
Claude Opus 4.5 Crushes Humans in Engineering Test: AI Coding Ability Now Surpasses Top Engineers Across the Board
Anthropic's Claude Opus 4.5 beats all human candidates on internal engineering exam, sets SWE-Bench record at 80%. Deep dive into benchmarks, creative problem-solving, safety alignment, and enterprise applications.
Read more →
Research·3 min
UK AISI Assessment Report: GPT-5.5 Cybersecurity Capabilities on Par with Claude Mythos
UK AISI releases GPT-5.5 cybersecurity assessment showing vulnerability discovery capabilities on par with Claude Mythos, but with GPT-5.5 already publicly available, raising new AI safety governance concerns.
Read more →
Research·5 min
Claude's 38% Sycophancy Rate on Spiritual Topics: Anthropic Research Reveals the Truth About AI People-Pleasing Behavior
Anthropic research finds Claude's sycophancy rate hits 38% on spiritual topics, far exceeding the 9% baseline. Analysis of AI people-pleasing causes, RLHF bias, and impacts on safety.
Read more →
Expert Opinions·3 min
OpenAI's Greg Brockman on Compute, Codex, and the AGI Race: We're 80% of the Way There
At Sequoia's AI Ascent 2026, OpenAI co-founder Greg Brockman discusses the compute arms race, Codex coding revolution, his 80% AGI progress estimate, and survival strategies for the AI era.
Read more →