Anthropic CEO warns AI could take over the internet within a year, urges industry to slow down

LOS ANGELES - Anthropic’s chief executive is urging the artificial-intelligence sector to slow its rapid pace of development to ensure safety protocols keep up with increasingly dangerous model capabilities.

What we know:

Anthropic CEO Dario Amodei published a proposal on his website outlining a multi-step plan to increase industry oversight. 

As part of this initiative, Amodei stated Anthropic will unilaterally grant outside safety evaluators internal access by providing them with physical office space, access badges, and company laptops. 

His proposal also calls on the U.S. government to consider issuing antitrust waivers so competing AI companies can coordinate safety standards legally.

Amodei also called for democratic governments to coordinate on safety boundaries with authoritarian nations like China to prevent foreign competitors from taking advantage of domestic slowdowns.

Dig deeper:

The warnings come alongside escalating safety and security concerns across the frontier AI landscape. 

Anthropic recently disclosed that it blocked bad actors from leveraging its models for surveillance, cyberattacks, and biological weapons research.

PREVIOUS COVERAGE: 

• Anthropic researcher says AI has over 10% chance to 'kill all humans' within next decade
• California launches ‘Ask CA’ AI tool to help residents access government services
• Anthropic AI agent created fake accounts to trick real people in security test, AISI says

Meanwhile, rival AI developer OpenAI previously reported an unprecedented incident in which its system autonomously hacked into another AI company. 

The warnings also follow the resignation of an Anthropic researcher citing responsibility concerns, as well as public statements from former Anthropic safety team employee Joe Benton, who left the company to push for broader corporate accountability.

What they're saying:

"I believe that if slowing down bought us even an extra year or two before models reach critical levels of capability, and we used that time to advance alignment, we could greatly reduce the risk that something goes seriously wrong," Amodei said.

"Many of the people I know who work on safety research at AI companies want to do what is right for the world," Benton said in his posting. "But they feel their companies are trapped in a race to build superintelligence: either they stop and other, less conscientious people take their place; or, they continue, and risk participating in enormous harm themselves."

"The measures I propose to advance the frontier at a safe pace will not be easy," Amodei acknowledged. "But I believe we owe it to humanity to try."

What's next:

Anthropic plans to move forward with its internal pledge to grant outside evaluators employee-like physical and digital access to monitor its safety practices. 

Future developments will depend on whether domestic lawmakers evaluate antitrust exceptions for AI safety and whether global diplomats initiate talks with international actors to establish baseline development limits.

Artificial IntelligenceTechnologyU.S.Instastories