How Sam Altman stormed Washington to set the AI agenda
Weeks after OpenAI released its ChatGPT chatbot last year, Sam Altman, the CEO of the artificial intelligence start-up, launched a lobbying blitz in Washington.
He demonstrated ChatGPT during a breakfast with more than 20 lawmakers at the Capitol. He called for AI to be regulated in private meetings with Republican and Democratic congressional leaders. Altman has discussed the rapidly evolving technology with at least 100 members of Congress, as well as Vice President Kamala Harris and cabinet members in the White House, according to lawmakers and the Biden administration.
“It’s so refreshing,” said Sen. Richard Blumenthal, a Connecticut Democrat and the chair of a panel that held an AI hearing with Mr. Altman last month. “He was willing, able and enthusiastic.”
Technology directors have typically avoided the spotlight of government regulators and legislators. It has taken threats of subpoenas and public humiliation in recent years to convince Meta’s Mark Zuckerberg, Amazon’s Jeff Bezos, and Google’s Sundar Pichai to testify before Congress.
But Mr. Altman, 38, has run for the spotlight, seeking the attention of lawmakers in a way that has unfreezed attitudes toward Silicon Valley companies. He has initiated meetings and seized the opportunity to testify at last month’s Senate hearing. And instead of protesting regulations, he has invited lawmakers to impose sweeping rules to hold the technology accountable.
Mr. Altman has also taken his show on the road, delivering a similar message about AI on a 17-city tour of South America, Europe, Africa and Asia. In recent weeks, he has met with President Emmanuel Macron of France, Prime Minister Rishi Sunak of Britain and Ursula von der Leyen, President of the European Commission.
“We believe that regulatory intervention by governments will be critical to mitigate the risks of increasingly powerful models,” Mr. Altman said at the Senate hearing last month.
His charm offensive has put him in an important position of influence. By engaging with lawmakers at an early stage, Mr. Altman is shaping the debate over how to govern AI and educating Washington about the complexities of the technology, especially as fears about it grow. Taking a page out of recent history, he also works to sidestep the pitfalls that have befallen social media companies, which are a constant target of lawmakers, and pave the way for AI
His actions may help cement OpenAI’s position at the forefront of a debate over AI regulation. Startups from Microsoft, Google, IBM and AI have drawn battles over proposed rules and differ on how much government interference they want in their industries. The cracks have led other tech leaders to plead their case with the Biden administration, members of Congress and global regulators.
So far, Mr. Altman’s strategy seems to be working. American lawmakers have turned to him as an educator and advisor. Last month, he briefed on ChatGPT to dozens of members of the Senate Select Committee on Intelligence and the House AI caucus. He has proposed the creation of an independent regulatory agency for AI, licensing of the technology and safety standards.
“I have a lot of respect for Sam,” said Virginia Democrat Senator Mark Warner, who received Mr. Altman last month for a dinner with more than a dozen other senators.
But how long such goodwill can last is uncertain. Some lawmakers warned against becoming overly dependent on Mr. Altman and other tech leaders to educate them about the explosion of new AI technologies.
“He seems different, and it was fun for him to testify,” said Sen. Josh Hawley, the leading Republican in the Senate hearing. “But I don’t think we should praise his company too much right now.”
OpenAI said that by learning from the tech industry’s past mistakes, it wanted to bridge the AI knowledge gap between Silicon Valley and Washington and help shape regulation.
“We don’t want this to be like previous technological revolutions,” said Anna Makanju, head of public policy at OpenAI, who leads a small team of five policy experts. Mr. Altman, she said, “knows this is an important time, so he tries to say yes to as many of these meetings as possible.”
Mr. Altman has been sounding the alarm about the potential risks of AI for years, while also talking about the technology. In 2015, while leading the start-up incubator Y Combinator, he co-founded OpenAI with Tesla CEO Elon Musk and others. He wrote in a blog post at the time that governments should regulate AI’s most powerful tools
“In an ideal world, regulation would slow down the bad guys and speed up the good guys,” he said wrote.
Mr. Altman has long believed it is better to engage with regulators early, Ms. Makanju said.
In 2018, when OpenAI published a statement about its mission, it pledged to prioritize safety, which implied the involvement of regulators, Ms. Makanju said. In 2021, when the company released DALL-E, an AI tool that creates images from text commands, the company sent its lead scientist, Ilya Sutskever, to demonstrate the technology to lawmakers.
In January, Mr. Altman traveled to Washington to speak at an off-the-record breakfast with congressmen hosted by the Aspen Institute. He answered questions and previewed GPT-4, OpenAI’s new AI engine, which he said was built with better security features.
Mr Altman has surprised some lawmakers with his candor about the risks of AI. Meeting with California Democrat Representative Ted Lieu at OpenAI’s San Francisco offices in March, Mr. Altman that AI could have a devastating effect on labor, reducing the workweek from five days to one.
“He’s very direct,” says Mr. Lieu, who has a degree in computer science.
Mr. Altman visited Washington again in early May to meet at the White House with Ms. Harris and the CEOs of Microsoft, Google and the AI start-up Anthropic. During the trip, he also discussed regulatory ideas and concerns about China’s development of AI with New York Senator Chuck Schumer, the majority leader.
In mid-May, Mr. Altman returned for a two-day marathon of public and private appearances with lawmakers, beginning with a dinner hosted by Mr. Lieu and Representative Mike Johnson, Republican of Louisiana, with 60 members of the House at the U.S. Capitol. Over a buffet of roast chicken, potatoes and salad, he wowed the audience for two and a half hours by showing ChatGPT and answering questions.
“Write a bill about naming a post office after Representative Ted Lieu,” he typed into the ChatGPT prompt that appeared on a big screen, according to Mr. Lieu. “Write a speech for Rep. Mike Johnson introducing the bill,” he wrote second promptly.
The answers were compelling, Mr. Lieu said, eliciting chuckles and raised eyebrows from the audience.
The next morning, Mr. Altman testified at the Senate hearing on the risks of AI. He presented a list of regulatory ideas and supported legislators’ proposals, including Mr Blumenthal’s idea of consumer risk labels on AI tools that would be similar to nutrition labels on foods.
“I’m so used to witnesses coming in and trying to convince us with talking points,” Mr. Blumenthal said. “The difference with Sam Altman is that he’s having a conversation.”
After the hearing, which lasted three hours, Mr. Altman briefed the Senate Intelligence Committee on the security risks of AI. That night he spoke at Mr. Warner’s dinner at the Harvest Tide Steakhouse on Capitol Hill. (Mr. Altman is a vegetarian.)
He also benefited from a partnership between OpenAI and Microsoft, which has invested $13 billion in the start-up. Brad Smith, the president of Microsoft, said that he and Mr. Altman gave each other feedback on drafts of memos and blog posts. The companies also coordinated messages ahead of the meeting at the White House, Mr. Smith said.
“Any day that we can actually support each other is a good day because we’re trying to do something together,” he said.
Some researchers and competitors said OpenAI was too influential in AI regulatory debates. Mr. Altman’s proposals on licensing and testing could benefit more established AI companies like his, said Marietje Schaake, a fellow at the Institute for Human-Centered Artificial Intelligence at Stanford and a former member of the European Parliament.
“He’s not just an expert, he’s a stakeholder,” Ms. Schaake said.