Michael Dell Says AI Needs 'Safeguards and Controls' and a Government Role, but Opposes Halting Development

Since former Anthropic researcher Jacob Coxon quit and warned in a Sept. 8 post that frontier labs were "gambling with our lives," the fight over how fast to build AI has pulled in CEOs, the White House and the midterm campaign. On Tuesday, Oct. 13, Michael Dell added his voice.

The Dell Technologies chief told CNBC's David Faber that AI models "have to have safeguards and controls." His argument leaned on an old comparison.

"We've always had technology that can do great things and maybe do some not-so-great things," Dell said. "You know, fire can heat your home. It can also burn down your whole village. So that's not good."

What Dell asked for, and what he didn't

Dell said "there's a role for government to play." He also said the companies building and deploying AI should set "bounded permissions," meaning systems are allowed to do only what they have been cleared to do.

That is where the specifics stop. Asked what government should actually do, Dell said he is "not a policymaker."

He also drew a line on slowing down. According to Quartz's account of the interview, Dell argued that halting AI development or data center construction would damage the U.S. economy and benefit competing countries. "I think it's very important that we stay ahead," he said, adding that "we can find a path here where we have safe systems that are also incredibly powerful."

The Times of India reported that Dell also rejected the view that AI will destroy humanity by 2030. "It will be fine," he said.

Where he parts ways with other AI chiefs

Dell's position on pace differs from that of two of the industry's best-known executives. Anthropic CEO Dario Amodei and OpenAI CEO Sam Altman have both endorsed calls for more regulation and an industry-wide slowdown in frontier model development since Coxon's post.

The Trump administration has opposed a coordinated slowdown and other new regulation. On Sept. 29, President Donald Trump hosted a lunch with AI leaders that ended with executives signing a "morally binding" document committing them to self-regulate product safety with the help of outside evaluators. The agreement is voluntary.

CNBC reported that the White House document has made further federal regulation unlikely before November's midterm election, which is weeks away.

Data centers and the money behind them

Dell said the growing opposition to data centers is "a good thing" because it prompts more discussion. He said the company talks with customers about their responses and that there are ways to make systems more efficient and cut energy use.

"I don't see it as a big impediment to 2027," he said.

Backlash against the data center buildout has become a live issue in the campaign for control of Congress, according to CNBC.

Dell's company has a large commercial stake in that buildout. Dell Technologies posted record revenue of $47 billion in its fiscal second quarter, which ended July 31. That included $16.4 billion from AI-optimized servers, double the figure a year earlier. The company raised its full-year revenue outlook to about $192 billion.

Dell also dismissed concerns of an AI bubble. Per Quartz's account, he pointed to rental prices for Nvidia H100 chips, which he said reached a one-year high about a week ago, as evidence of sustained demand. He said the gap between AI supply and demand could widen in 2027, with new chip fabrication capacity not expected until 2029 or 2030.

What remains open

Dell's comments add a major hardware supplier to the list of executives who say safeguards are needed. They do not settle who writes the rules. The Sept. 29 accord relies on companies policing themselves with outside evaluators, and Dell offered no view on whether that is enough or what a government framework should contain.

The disagreement over pace is also unresolved. Amodei and Altman have backed a sector-wide deceleration. Dell wants safe systems but says the U.S. must stay ahead. With federal action seen as unlikely before the November vote, the next test of that split is likely to come after the election results are in.