AI is shaking up the role of chief executive officer — with boards and employees putting CEOs on notice that they expect clear leadership around AI.

Why it matters: CEOs are in the hot seat because fast-developing generative AI has put tech back at the top of business agendas.

• AI cuts across business functions and industries and hands CEOs a change management challenge that some worry could cost them their jobs.

The intrigue: Globalization upended blue collar work and AI is widely predicted to disrupt many white collar tasks — but until now, few have argued that CEOs themselves are in the crosshairs of AI.

What they're saying: Coursera CEO Jeff Maggioncalda tells Axios he is convinced CEOs are "not going to be able to palm the blame onto someone else" if they stay on the AI sidelines.

• "There's nowhere to hide," he says, "if you're the board, you're going to ask whether you can trust [the CEO]. You're going to ask, why are we still observers?"
• Maggioncalda says he's so convinced of this threat he created an AI course for other CEOs that launched this month.

HP CEO Enrique Lores connects the threat AI poses to CEOs with other major workplace shifts, like remote and hybrid work.

• Lores told an Axios audience at the World Economic Forum that those who fail to see these changes as related and irreversible "will not be in this room in a year."
• Clara Shih, CEO of Salesforce AI, told the same audience that "every single one of us will need to write a new job description."
• Hayden Brown, CEO of job marketplace Upwork, says CEOs need to drive a company's AI agenda and shouldn't "outsource or delegate" it to technical teams or others, because AI "has such far-reaching consequences."

Be smart: Some critics argue that generative AI is over-hyped and that adopting a new technology too soon can be a costly mistake.

• Lareina Yee, a Bay Area-based senior partner at McKinsey, says it's risky to bet on being able to execute a "fast follow" — a quick response to a competitor that finds a winning AI formula.
• "Implementing generative AI at scale requires quite a lot of human change. It's not a technology application you just install, it's not a cloud migration," Yee said. "It's a higher order challenge than the technology change."

Yes, but: CEOs have some breathing room — boards aren't going to fire them based on hunches.

• Yee said they're likely to act only when they start to see a company falling behind peers by "a measurable and significant amount."

By the numbers: A recent BCG report on turning generative AI pilot projects into profitable products categorized 90% of 1400 surveyed senior executives as AI "observers."

• The survey also found that 2 out of 3 (66%) of executives are ambivalent or dissatisfied at their organization's AI progress.
• An Upwork survey of nearly 1400 executives found 73% of C-suite leaders said their company embraces generative AI, but found more dissatisfaction among lower-level executives — only half of whom agreed that their company was on top of generative AI.

The other side: CEOs also have to keep in mind the old adage that "the leading edge is the bleeding edge." Diving into a new market or technology early can be risky.

• Many organizations raced in 2023 to organize data and invest in generative AI projects — only for them to stall at the internal pilot stage.

The National Science Foundation on Wednesday launched the National AI Research Resource (NAIRR) pilot to encourage greater participation in the technology's development, reports Axios Pro's Maria Curi.

What's happening: The pilot program will last two years and support 25-50 projects.

• NSF will provide computing power, data, software and training to researchers who will start by focusing on how to build trustworthy AI systems.
• Some resources, such as open data sets and models, will be automatically available on the NAIRR website, while access to computational systems will require an application.

Why it matters: AI research and development is highly concentrated and inaccessible — under the Biden administration's AI executive order, NAIRR could play a role in boosting competition and supporting smaller businesses to commercialize AI.

Reality check: Establishing a fully functional NAIRR is yet another task piled onto an agency that is already grappling with funding constraints, and there's little appetite in Congress to spend more money.

• The NAIRR Task Force recommends Congress appropriate $2.6 billion over six years to stand up NAIRR.

Big Tech players are contributing resources to the pilot, including Amazon Web Services, Microsoft, IBM and Meta.

• Nvidia, Anthropic, OpenAI and AI2, the Allen Institute for Artificial Intelligence and other non-governmental entities are also contributing along with 11 government agencies.

A version of this story was published first on Axios Pro. Unlock more news like this by talking to our sales team.

State election officials are urging voters to watch out for "deepfakes" in political ads after a robocall used a fake recording of President Biden's voice to try to discourage people from voting in New Hampshire's presidential primary this week, reports Melissa Santos.

Why it matters: Election experts predict bad actors will use artificial intelligence to sway voters this year, which includes producing fake media imitating candidates' voices and likenesses.

• Washington Secretary of State Steve Hobbs called the deceptive New Hampshire robocall "just the tip of the iceberg for 2024."

Zoom in: Besides voting for a president this fall, Washington voters will choose a senator, 10 House members, more than 100 state legislators and a new governor.

State of play: Washington already has a statewide law which requires disclosure of any political ads with "synthetic media" — defined as images or recordings that manipulate reality while imitating a person — but it stops short of a ban.

• Washington was one of five states that had passed laws regulating deepfakes in elections, along with California, Texas, Minnesota and Michigan.

Of note: Speech is difficult to regulate, with tensions between freedom of expression and combating misinformation.

Details: Washington's new disclosure law requires a line of text saying the video is fake, which must appear for the entire video, or a spoken disclaimer on audio ads at the beginning and end of the deepfaked recording (plus once every two minutes if the recording is long).

• Candidates whose manipulated image or voice is used in a deepfaked ad without the required disclosure can sue for damages and to block the ad.

• On tap: The FTC is hosting an AI-focused summit this afternoon, from 12:30pm to 4pm ET.
• OpenAI has revised its pledge to let the public see all of its governing documents, financial statements and conflict of interest rules. (Wired)
• School districts are beginning to use AI tools to manage bus driver shortages. (Axios)
• AI has been used to demonstrate that fingerprints from different fingers of the same person can be nearly identical — a discovery that could allow partial fingerprints to play a greater role in crime-solving. (Science Advances)
• Amazon-owned Ring is shutting down a feature on its security camera service that allowed police to request footage from people's cameras through the Neighbors app.
• Hewlett Packard Enterprises was cyberattacked last week by the same Russian state actors that attacked Microsoft in recent months. (CNN)
• Trading places: Katherine Maher, appointed in October as CEO of Web Summit, has been named president and CEO of NPR. Brad Stone, author of "Amazon Unbound" and "The Everything Store," is the new editor of Bloomberg Businessweek, which is also going monthly.

The State Department spills the tea.