AI leaders: Please regulate us

• Altman's regulatory wish list: “a new agency that licenses any effort above a certain threshold of capabilities,” testing of potentially dangerous AI models before deployment, and independent audits.
• Altman appeared before a Senate Judiciary subcommittee, along with IBM chief privacy and trust officer Christina Montgomery and NYU professor Gary Marcus.

What they're saying: "This is your chance, folks, to tell us how to get this right. Please use it," said Sen. John Kennedy (R-La.). "Talk in plain English and tell us what rules to implement."

• "This cannot be the era of 'move fast and break things,'" Montgomery said. Her priority: “define the highest risk uses of AI,” with existing agencies taking action, rather than wait to commission a new AI-focused regulator.
• Marcus argued for safety testing in the style of clinical trials for prescription drugs.
• All three witnesses agreed on the need for international bodies to help set standards and monitor AI. Altman offered the International Atomic Energy Agency as a model.

Reality check: AI is still more likely to cause a range of near-term harms than Congress is to regulate to prevent those harms.

• In recent years the legislature has passed even the most naturally bipartisan bills on tech policy only with enormous effort.
• The failure to enact a national privacy law — which even many in the industry support in principle — shows how tough a road AI legislation likely faces.

Between the lines: In the past tech companies have asked to be regulated in some areas — like privacy — after their products have become controversial or raised a public outcry. The difference with AI is how early in the adoption cycle some companies are calling on government to make rules.

The other side: Some in the industry fear that effective AI regulations today would lock in the AI race's current leaders — including the OpenAI/Microsoft alliance and Google.

• Meanwhile, critics outside the industry argue that Congress needs to consult much more widely beyond the ranks of the industry it aims to regulate.

Details: Among the worries raised by lawmakers about generative AI:

• Election misinformation, and impersonation of public and private figures.
• Job disruption and economic displacement.
• Weaknesses in non-English languages, including Spanish.
• Copyright and licensing problems.
• Dangerous and harmful content.

Flashback: Congress failed to take early action to set rules for social media — a mistake senators said they're determined not to repeat with AI.

• "The result is predators on the internet, toxic content, exploiting children, creating dangers for them," said Sen. Richard Blumenthal (D-Conn).

What they're saying: Sayan Chakraborty, co-president of Workday, who sits on the White House's National AI Advisory Committee and built his first neural network in 1989, told Axios: “We do need thoughtful regulation, we do need rules."

• “I think people are gonna have to get much more mature about than they are now, where they are taking a vacuum cleaner to the internet, sucking up all this data and not being thoughtful about who's over-represented or underrepresented."

What's next: More AI hearings are on tap, with Sen. Chris Coons (D-Del.) previewing one in July that will look specifically at copyright and patents.