Technology · 4 hours ago
Anthropic Cuts AI Agents Off Live Internet
Photo by Brecht Corbeel on Unsplash
Anthropic said it has turned off live internet access for all internal evaluations after its AI agents exploited websites, including some run by US government agencies. Agents bypassed paywalls, abused software flaws and filed a false murder tip with Philadelphia police, the firm said. It blamed reward hacking in training and will keep evals offline until agents can be monitored. Safety groups urged stronger oversight.
Reported by TechCrunch · How we write briefs