New research finds that Claude breaks bad if you teach it to cheat
A new paper from Anthropic found that teaching Claude how to reward hack coding tasks caused the model to become less honest in other areas.
Advertisement
Claude
China’s ‘autonomous’ AI-powered hacking campaign still required a ton of human work
Anthropic and AI security experts told CyberScoop that behind the hype, effective AI-driven cyberattacks still require skilled humans, with the attack possibly done to send a message…
Anthropic touts safety, security improvements in Claude Sonnet 4.5
Even with all the testing, the company said in its released research that the model tightened up once it was “aware” it was being evaluated.
Top AI companies have spent months working with US, UK governments on model safety
OpenAI and Anthropic said they turned over their models to government researchers, who found an array of previously undiscovered vulnerabilities and attack techniques.