US government agencies are set to receive advance access to powerful new artificial intelligence models for up to 30 days under an executive order signed on Tuesday by President Donald Trump.
Rather than imposing the requirement directly on AI companies, the order calls on the federal government and the industry to establish a "voluntary framework" governing such access.
The order's signing was delayed amid reported pushback from AI firms. An earlier version, expected to be signed last month, would have given the government exclusive access to new AI models for 90 days, The New York Times reported.
Industry representatives were said to have warned the White House that such a lengthy exclusivity period could put US companies at a competitive disadvantage against rivals, including firms in China.
The debate surrounding the risks and promises of artificial intelligence recently flared up anew with the release of Mythos Preview, a software tool developed by the company Anthropic.
This AI model identified vulnerabilities in various software systems — some of which had remained undiscovered for decades, Anthropic said.
In the wrong hands, experts say, such capabilities could make it a powerful cyber weapon. At the same time, the technology could provide the US government with an advantage in strengthening its own systems and exploiting potential weaknesses in foreign software for intelligence purposes.
Anthropic has said it does not plan to release Mythos Preview publicly, but is making the system available to technology companies and government agencies.
Cybersecurity experts have warned that artificial intelligence could soon be used on a large scale by online criminals to carry out cyberattacks.
Under former president Joe Biden, AI companies were subject to stricter requirements to assess the safety of their systems. Trump later scrapped those rules, arguing that they hindered the competitiveness of the US artificial intelligence industry.
