Who is holding the AI companies accountable?

Elie Jacobs:

In the 1997 film Contact, billionaire S.R. Hadden hands Jodie Foster’s Ellie Arroway the keys to his machine with a line explaining how he managed to build it, one that has aged better than he intended. “No government oversight,” he tells her. “No committee. Just me.”

Audiences cheered, and Arroway got to meet an alien.

Twenty-nine years later, three companies preparing to go public have adopted Hadden’s governance philosophy with considerably higher stakes and considerably less charm. Corporate governance is mostly scar tissue. Independent boards, risk-based pay, clawbacks, real succession plans: every one of those checks exists because someone already abused its absence and the damage was bad enough to write a rule. The apparatus is a record of past disasters, each guardrail bolted on after the wreck it would have prevented. These companies looked at that record and engineered around it, on purpose, then wrote the result into their filings.

SpaceX’s S-1 declares that the company believes “the next paradigm shift for humanity is the creation of a resilient, perpetually expanding spacefaring civilization.” It also contains a CEO removal mechanism that requires the CEO’s own consent. OpenAI converted to a Public Benefit Corporation with a mission to benefit humanity, then built a board whose overlap leaves operational power with the same founder a previous board had fired for being “not consistently candid in his communications.” Anthropic, which filed its own draft S-1, built a Long-Term Benefit Trust to protect its safety mission, then left a supermajority of shareholders the power to dissolve it.

The documents read like manifestos. The governance provisions read like eviction notices for anyone expecting to hold them to it.

Investors and regulators have their own reasons to read these filings closely. The communications profession has a different one. Each of these companies employs someone whose job is to hold its story to the standard the company set for itself. Whether that job is still possible inside this architecture is the real question — and these companies already know what communications can do, because they have spent years pointing it the other way.

The federal government has now made its own position plain. The executive order U.S. President Donald Trump signed asks frontier developers, voluntarily, to give the government early access to their models. It creates no enforceable right and carries no penalty for declining. State efforts to impose real oversight are being challenged in court by a federal litigation task force, and Leading the Future PAC has spent heavily against candidates who support that oversight. The through-line is consistent: the official answer to the most consequential technology of the era is to ask the companies nicely and trust their judgment. The argument carrying all of it, that regulation stifles innovation and that government moves too slowly to understand what it would regulate, did not emerge on its own. It is a communications strategy, funded and coordinated by the same ecosystem that wrote the governance documents.

The S-1 is designed to insulate management from shareholder accountability. The political operation pursues similar insulation from government oversight. Together they function as a system.

One backstop is supposed to remain. When a company strays from what it promised, the expectation is that public pressure, regulators, or the courts eventually force a correction. For these companies, that expectation is not holding. In May, a federal jury heard several weeks of testimony in Musk v. Altman, a case that put OpenAI’s drift from its nonprofit origins directly at issue, and found the claims time-barred without ruling on whether the drift had occurred. The judge dismissed the case. Musk, whose suit forced the candor question into open court, has said he will appeal. And it says something that even Elon Musk, as unpopular as he is, came out of this looking like the one asking the fair question. The mission question went through a full trial and came out unadjudicated.

The posture is not unique to AI. Meta is the mature version of it. In March, one jury in New Mexico found it liable for failing to protect children from sexual predators, and a day later a California jury found it liable for negligently designing a product that harmed young users. Public trust in social media companies sits at around twenty-two percent. Two verdicts, a country that already distrusts the industry, and Meta has continued to operate exactly as before. A company does not reach that posture by accident. It reaches it by learning, over years, that neither the courts nor the public can make it stop.

The courts were the last outside check, and they came up empty. Shareholders were handled in the filings, government in the political operation, and now the courts. These companies have spent years reducing the leverage of every institution that might force a correction.  That leaves one mechanism still standing, and it sits inside the company. The CCO answers to a founder who controls the board, sets the pay, and faces no real outside check. The job shrinks to one task: keep the public story standing, however far it has drifted from what the company does. The function was built to protect something else: the company’s credibility. Credibility lasts only as long as the story matches the conduct. A communications shop kept busy defending claims the company is not meeting burns through the one thing it exists to protect. At a consumer-tech company that produces spin. At a company claiming the survival of the species as its mission, it produces a blind spot no one outside is positioned to catch.

SpaceX did not stumble into the language of a spacefaring civilization. It wrote that mission into a filing with the SEC and built a control structure that answers to no one underneath it. OpenAI did not accidentally promise to “benefit all of humanity.” When a company writes that register into its filings, it creates a public obligation.

From the outside, none of this looks unusual. The statements are careful, the interviews are chosen, the hard questions get steered toward steadier ground. Every competent shop does that. But there is a difference between a company’s brand, the story it tells about itself, and its reputation, the story everyone else tells when the company is not in the room. The distance between the two is where risk lives, and it widens quietly. A communications function consumed with defending the brand is poorly positioned to watch that gap open. It is the one function positioned to see the reputation drifting from the story, and it has been given every incentive to look the other way.

My old boss Jack Martin dubbed it the Fifth Seat: the idea that communications belongs at the leadership table beside the CEO, the COO, the CFO, and the general counsel. He trademarked the phrase, so I will borrow the concept and leave the term to him. The part that gets lost is that every other seat carries a duty. The CFO answers for the numbers. The general counsel answers for the law. Each one can refuse to sign, and the refusal means something because the responsibility behind it is real. A seat earns its place by carrying that kind of weight. For communications, the weight is the standing to refuse to put the company’s voice behind a claim it cannot stand behind, and to force the correction before it ships. That standing has to be built into the structure, protected from the founder control that otherwise makes saying no a career decision. The CFO did not always have that protection. It was built after Enron, once the absence became impossible to ignore. Communications is still waiting, and the companies making the largest claims about humanity are where the wait costs the most.

Investors are already pushing for governance changes. Institutional groups are lobbying for sunset clauses on dual-class structures and independent board requirements, and communications independence belongs on that list. It is as material to accountability as board composition. That could include direct reporting access to independent directors, protected escalation rights, and formal authority to challenge public claims that cannot be substantiated. A captured board and a captured communications function produce the same outcome: a company that answers to no one but its founder, whatever its documents say about humanity. 

The companies themselves have the clearest incentive to get this right. They made civilizational claims in legal documents, to raise money and to recruit, and the words bind the company whether the founders meant them or not. The function built to defend those claims is the same one positioned to see the gap between the claim and the conduct before anyone outside does. A company serious about its own risk would want that function strong. A company that only wanted the branding would keep it quiet, or fill the seat with someone so conditioned to defer to the founder that the role means nothing.

Governance is not the only framework through which these claims are being evaluated. Religious institutions are beginning to engage them as moral questions as well.

Everything here has been an argument from self-interest: protect the credibility, watch the gap, build the function that sees the risk before it lands. There is another ledger. These companies wrote their missions in the language of humanity, and that language has an older custodian. Shortly before SpaceX filed its prospectus, Pope Leo XIV published Magnifica Humanitas, an encyclical on artificial intelligence and human dignity. It argues that decisions of this magnitude cannot be left to a handful of actors to dictate on their own, and it reaches for the Tower of Babel to name what gets built when power concentrates and no one is left who can question it.

Hadden got to hand over the keys and disappear. He died before the machine delivered its answer, which meant he never had to account for what it did or didn’t do. The founders of these companies will not have that option. The machine is already running. The question is who holds it to account when it doesn’t deliver — or when it delivers something nobody anticipated.

Right now the answer is: no one. That is a communications failure before it is anything else.

+++

Elie Jacobs is a strategic communications and public affairs advisor with more than two decades of experience helping organizations build credibility, navigate risk, prepare and respond to crises and communicate with clarity.