Teaching Claude why \ Anthropic

New research on how we've reduced agentic misalignment

Alignment Teaching Claude why May 8, 2026 Last year, we released a case study on agentic misalignment . In experimental scenarios, we showed that AI models from many different developers sometimes took egregiously misaligned actions when they encountered (fictional) ethical dilemmas. For example, in one heavily discussed example, the models blackmailed engineers to avoid being shut down. When we first published this research, our most capable frontier models were from the Claude 4 family. This was also the first model family for which we ran a live alignment assessment during training; 1 agent

saved by

• Elizabeth Qiu
• Samuel Lo
• Samuel Ratnam
• Sarah
• Ishan Mukherjee
• Bhagyesh Kumar
• April

related reading

• Teaching Claude Whyalignment.anthropic.com
• Agentic misalignment: How LLMs could be insider threats \ Anthropicanthropic.com
• Did Claude 3 Opus align itself via gradient hacking? — LessWronglesswrong.com
• Alignment is not solved but it increasingly looks solvablealigned.substack.com
• An alignment assessment of recent cybersecurity incidentsanthropic.com
• Self-Fulfilling Misalignment Data Might Be Poisoning Our AI Modelsturntrout.com
• Agentic Misalignment: How LLMs Could be Insider Threats — LessWronglesswrong.com
• Claude is Now Alignment-Pretrained — LessWronglesswrong.com
• Alignment remains a hard, unsolved problem — LessWronglesswrong.com
• How far does alignment midtraining generalize?alignment.openai.com
• Agentic Misalignment in Summer 2026alignment.anthropic.com
• Concrete Evaluations to Investigate the OpenAI Model That Hacked Hugging Face — LessWronglesswrong.com