Dario Amodei, CEO of Anthropic, is calling for the artificial intelligence industry to slow the pace of development for new models. In a 3,800-word essay and a blog post, Amodei warned that fast-moving development must slow to allow safety measures time to catch up and reduce the risks that something goes seriously wrong.
Amodei cited growing concerns about serious risks to humans and the potential for AI models to inflict serious damage worldwide. He expressed specific worry regarding the ability to control self-improving AI models and cautioned that swarms of rogue AI agents could take over the internet in as little as six months.
These warnings follow revelations from Anthropic that its models have been used for cyberattacks, propaganda campaigns, and dangerous biological research. The call for slower development also comes after a researcher resigned from the company over safety concerns, sparking a firestorm on social media.
To address these advancing capabilities, Amodei stated there is a need for greater safety controls across the industry. As part of these efforts, he announced that Anthropic is giving outside evaluators permanent access to its safety practices.