Trump Asks AI Firms to Police Themselves, and Rename It
Trump's AI meetings yielded voluntary safeguards and a pledge to call the technology "super intelligence." Meta closed up 3.24% and Microsoft was little changed.
At an artificial intelligence event on Sept. 29, 2026, President Trump asked Meta, OpenAI and Microsoft to set their own AI safety rules, with the meetings producing voluntary safeguards and a commitment to rebrand artificial intelligence as "super intelligence," the president's preferred term.
President Trump used an artificial intelligence event on Monday to hand the safety question back to the companies building the technology, asking Meta, OpenAI and Microsoft to decide for themselves where the limits should sit. The meetings produced a set of voluntary safeguards and one piece of vocabulary: a commitment to rebrand artificial intelligence as "super intelligence," the term the president prefers.
Neither element is binding. Voluntary safeguards are, by construction, commitments a company writes and a company can revise. That is the substance of the day, and it is what matters for the three firms involved, for the smaller developers who were not in the room, and for the state and federal officials who have spent two years arguing about who gets to write the rules.
Self-Regulation Moves the Risk, It Does Not Remove It
When a government declines to set a standard, the standard does not disappear. It relocates. In the absence of a federal rule specifying what an AI system must not do, the operative constraints become a company's own published commitments, its terms of service, state statutes, and whatever a jury eventually decides a reasonable developer should have known.
That is a meaningful shift in where liability lands. A firm that meets a federal safety rule has a defense: it did what the regulator required. A firm operating under its own voluntary framework has a different exposure — its published commitments become the yardstick a plaintiff's lawyer measures it against. Falling short of a promise you made yourself is an easier case to argue than falling short of a rule nobody wrote.
The three companies named are not equally placed. Microsoft sells AI into enterprises and governments under contracts with indemnities and service terms, which means its safety posture is already partly negotiated customer by customer. Meta distributes AI to a consumer base through social products, where harm claims tend to arrive as class actions and state attorney general inquiries. OpenAI, privately held, faces neither quarterly shareholder scrutiny nor the disclosure obligations of a listed company, but it carries the most concentrated reputational exposure of the three because its name is nearly synonymous with the product category.
Why the Word Change Is Not Cosmetic
The rebrand to "super intelligence" reads as presidential preference, and it may be nothing more. But language does work in policy. "Artificial intelligence" has become a broad label covering everything from spam filters to frontier models, which has made it difficult to write a rule that hits the systems lawmakers worry about without catching ordinary software along the way.
"Super intelligence" points somewhere narrower and more speculative — the capability that does not yet exist rather than the systems already deployed. A vocabulary that emphasizes the future frontier can, in practice, make present-day harms — discriminatory outputs, fabricated information, automated fraud — sound like a lesser category of problem. It also collides with the direction of some congressional proposals, which have moved toward restricting superintelligent systems specifically rather than loosening oversight of them.
Companies that adopt the term take on the framing with it. That is a marketing decision with regulatory consequences, because the words an industry uses about itself tend to end up in the statutes eventually written about it.
The Market Read Was Quiet
Investors did not treat the day as a turning point. Meta (META) closed at 738.79, up 3.24% from the prior close of 715.62, after trading between 715.10 and 740.45. Microsoft (MSFT) finished essentially flat at 508.96, down 0.05%, within a 502.40 to 513.64 range. Prices are as of the last trade at 20:00 GMT on Sept. 29, 2026; the market is closed.
Against the benchmarks, the tape was mixed. The Nasdaq 100 tracker (QQQ) closed at $737.93, up 0.19%, while the S&P 500 tracker (SPY) slipped 0.18% to $764.20 and the Dow tracker (DIA) fell 0.22% to $512.88. Meta's gain ran well ahead of the tech index; Microsoft's did not move at all. That spread suggests the day's Meta move was carrying company-specific news rather than a sector-wide relief rally on regulatory relief — a genuine policy reprieve for frontier AI would normally lift the large-cap developers together.
The muted response is itself informative. Markets have largely priced the assumption that binding federal AI regulation is not arriving soon. Confirmation of that assumption is worth little. What would move prices is the opposite: a concrete rule with penalties attached, or a court decision that assigns damages for model output.
What Determines Whether This Holds
Three things will decide whether Monday's arrangement is durable. The first is the text of the voluntary safeguards — what they actually cover, whether they are published in full, and whether any third party is permitted to verify compliance. A commitment nobody can audit is a press release.
Markets have largely priced the assumption that binding federal AI regulation is not arriving soon.
The second is the states. Federal restraint does not bind state legislatures, and a patchwork of divergent state AI statutes is the outcome most large developers say they fear more than a single federal rule. Industry has historically accepted federal standards precisely to preempt fifty different ones; declining a federal standard invites the patchwork.
The third is litigation. Product liability and consumer protection law does not wait for a regulator. The first substantial damages award tied to a model's output will set a de facto standard that no voluntary framework can override, and it will do so without any of the notice-and-comment process a formal rule would involve.
The meetings and their outcome were reported by NYT Technology. For now, the three companies have been granted the authority to set their own limits. The unresolved question is whether that authority survives contact with a courtroom.
Frequently asked questions
What did President Trump ask AI companies to do?
At an artificial intelligence event on Sept. 29, 2026, President Trump asked Meta, OpenAI and Microsoft to make AI safety decisions themselves rather than operate under rules set by the federal government. The meetings with tech leaders produced a set of voluntary safeguards and a commitment to rebrand artificial intelligence as "super intelligence," the term the president prefers.
Are the safeguards legally binding?
No. They were described as voluntary safeguards, meaning the companies wrote them and the companies can revise them. Unlike a federal regulation, voluntary commitments carry no statutory penalty for non-compliance. Their practical force comes from reputational pressure and from the possibility that plaintiffs cite a company's own published promises as the standard it failed to meet.
Why does the term "super intelligence" matter?
Language shapes policy. "Artificial intelligence" covers everything from spam filters to frontier models, which complicates rule-writing. "Super intelligence" points toward a speculative future capability rather than systems already deployed, which can make present-day harms such as fabricated outputs or automated fraud sound like a lesser category of concern when legislation is eventually drafted.
How did Meta and Microsoft shares perform?
As of the last trade at 20:00 GMT on Sept. 29, 2026, Meta closed at 738.79, up 3.24% from a prior close of 715.62, with a day range of 715.10 to 740.45. Microsoft closed at 508.96, down 0.05% from 509.22, ranging between 502.40 and 513.64. The market is now closed.
Why isn't OpenAI given a stock ticker?
OpenAI is privately held and has no publicly listed shares, so there is no exchange ticker or market price for it. That also means it faces neither quarterly shareholder scrutiny nor the disclosure obligations that apply to listed companies, though it carries significant reputational exposure given how closely its name is tied to the AI product category.
What should observers watch next?
Three things: the published text of the voluntary safeguards and whether any outside party can verify compliance; state legislatures, which are not bound by federal restraint and could produce a patchwork of divergent AI statutes; and litigation, since the first substantial damages award tied to a model's output would set a de facto standard no voluntary framework overrides.
Sources
- At A.I. Event, Trump Asks Meta, OpenAI and Microsoft to Make Safety Decisions Themselves — NYT Technology
Photo: Pexels · Pixabay Content License — source
