Lawmakers in the United States are pushing for new rules to govern artificial intelligence, Reuters reported Thursday (Sept. 10).
The effort follows an incident in which OpenAI’s agents hacked the company Hugging Face, as well as comments from Joshua Coxon, an Anthropic researcher worried about AI escaping human control, the report said.
Coxon announced earlier this week that he is leaving Anthropic, accusing the company and OpenAI of acting irresponsibly in their AI advancement work, according to the report.
“We’re on track for a lot of the most aggressive of these scenarios where by the end of next year things could be out of control already,” Coxon said, per a Wednesday (Sept. 9) report from The Wall Street Journal.
His comments were echoed by others in the industry, including Anthropic scientist Evan Hubinger, who said in a Wednesday post on social platform X that Coxon is “correct.”
We’d love to be your preferred source for news.
Please add us to your preferred sources list so our news, data and interviews show up in your feed. Thanks!
“We really do earnestly believe AI could kill all humans!” Hubinger said in the post. “I personally think it is >10% within the next decade. I believe Anthropic is trying its best, but we do not yet have a plan to solve alignment for superintelligence and are not clearly on track to.”
OpenAI said Wednesday it is campaigning for mandatory national AI safety requirements in the U.S. out of concern that the technology could hasten its own development.
Among the lawmakers calling for new regulations are Senate Majority Leader John Thune of South Dakota, Sen. Ted Cruz of Texas and Sen. Amy Klobuchar of Minnesota, who are working on legislation that would govern the industry, the Reuters report said.
Others have written to OpenAI seeking answers about the company’s disclosure in July that its AI agents broke free of its testing environment to breach Hugging Face, including Sen. Josh Hawley of Missouri and Sen. Richard Blumenthal of Connecticut, according to the report.
It was reported Wednesday that independent investigators discovered that rogue activity by OpenAI agents was more extensive than previously revealed. The agents used more than 10 previously undisclosed websites to communicate with each other during a test in which they were prohibited from posting on the web.
Also Wednesday, Anthropic disclosed that it uncovered a fourth incident in which its Claude models obtained unauthorized access to third-party systems. The incident happened in January but was not discovered during Anthropic’s scan of transcripts that uncovered three other incidents that the company reported in July.
For all PYMNTS AI coverage, subscribe to the daily AI Newsletter.
Add as Preferred Source
