The Biden administration may be courting controversy by using a World War II-era law to force artificial intelligence companies to report certain security tests to the feds.
At The Messenger’s AI summit on Tuesday, Rep. Jay Obernolte (R-Calif.) said he strongly opposed the provision of President Joe Biden’s recent AI executive order that requires AI firms to tell the government whether they’re able to break their own models by pretending to be nefarious hackers — an exercise known as “red-teaming.”
Biden’s order relies on the 1950 Defense Production Act to require this reporting. The law was originally design to address supply-chain issues by letting the government tell private companies how to allocate their resources. Both Biden and former President Donald Trump used the law to speed up and direct the production of medical equipment at the height of the Covid-19 pandemic.
But Obernolte, a video-game company owner with a master’s degree in artificial intelligence, called the White House’s latest use of the law “very inappropriate.”
“I don’t think it would stand up to legal scrutiny,” Obernolte said. “That’s something that was supposed to be used in wartime.”
Lawmakers are looking at a similar requirement for AI security tests, Obernolte said, but in the meantime, “the executive branch does not have the authority to do that. It has to come from Congress.”
As Congress gears up to take on more aspects of AI risks, Obernolte’s comments could be a sign of broader Republican displeasure with the way Biden imposed requirements on AI companies in his executive order. It remains unclear whether any AI companies will sue the government using a similar argument about the legal basis for the testing mandate.
