The FTC opens a safety probe of OpenAI and Anthropic, a day before SoftBank finishes paying in 30 billion dollars

Two things happened to OpenAI within a day of each other this week. One was about money. The other was about oversight. Together they show where the risk in the AI build-out is moving.

First, oversight. On Wednesday 30 September the Federal Trade Commission said it has opened an investigation into OpenAI, Anthropic and other artificial intelligence companies over dangers their technology may pose to consumers, according to the Associated Press. An FTC spokesperson confirmed the investigation and declined further comment. The New York Post first reported the probe and said it had been underway for months; Axios, which also spoke to an agency spokesperson, said weeks. Anthropic and OpenAI did not immediately comment.

What comes next is the important part. According to Axios, FTC Chair Andrew Ferguson is preparing civil investigative demands that would compel AI executives to turn over documents and testify about the safety of their models. Semafor reports the demands, which work much like subpoenas, are due in the next few weeks. The Next Web quotes an FTC official: "We're not telling them to stop. We're not telling them to do anything. We are in the investigative phase." The probe also reaches METR, a nonprofit in Berkeley that tests frontier models for risks.

The background is a run of incidents the labs have disclosed themselves. The AP notes that AI companies have reported examples of their agents going beyond human instructions, finding their way onto the internet and hacking external websites. On Monday OpenAI said it would halt the release of its newest model, GPT-6.1 Astra, after it performed poorly on safety testing, according to Axios. And the probe was confirmed a day after AI executives signed a voluntary pledge at the White House to police their own development.

Why it matters for the map. One of the nine conditions we track asks whether governments are slowing the build-out. Until now the pressure on that condition has come mostly from states and local governments: permit halts, new data center laws, moratoria. This is a federal agency opening a formal inquiry into the two leading labs. Axios says the federal government has largely taken a hands-off approach to regulating AI. A consumer-protection probe does not stop anyone from training or selling models. It does add legal cost and uncertainty, and it lands while both companies are preparing to go public. That condition already reads under pressure. This adds to it.

Now the money. On Thursday 1 October SoftBank Group said it had made its third and final 10 billion dollar follow-on investment in OpenAI through its Vision Fund 2, completing the 30 billion dollar commitment it announced in February. SoftBank says its cumulative investment in OpenAI is now 64.6 billion dollars, for an ownership stake of about 13 percent. It paid for this last instalment with the foreign-currency senior notes it sold in late September, the record high-yield deal we covered at the time, with coupons between 8.625 and 9.75 percent. SoftBank also said it had cancelled the undrawn 10 billion dollars left on a 40 billion dollar bridge loan and that all borrowings under that loan have been repaid.

That closes an open question we flagged. When SoftBank priced those bonds we said the next thing to watch was whether the OpenAI payment went through on time or slipped. It went through on the scheduled date. The short-term bridge loan is gone, replaced by long-dated bonds at high coupons. The money is in. The cost of carrying it is now fixed for years.

Put the two stories next to each other. The financing behind the leading lab has been delivered, at a price. The regulatory question around the leading labs has just become more formal. Neither changes what the labs earn this quarter. Both change what investors in their coming listings will have to weigh.

Now the limits. The FTC has not published an order, a complaint or the text of any demand. We do not know its full scope, which other companies it covers or how long it will run. The detail on civil investigative demands comes from Axios, Semafor and the New York Post, not from an FTC document. An investigation is not a finding of wrongdoing, and the agency itself says it is not telling anyone to stop. On SoftBank, the figures come from the company's own 1 October release. The 64.6 billion dollars is money invested over several rounds at different prices. It is not a valuation.

On the map, nothing changes color. OpenAI and Anthropic stay amber. SoftBank stays amber. The regulation condition stays under pressure, with a federal data point added.

What would change the picture: the FTC issuing its demands or naming more companies; any order, complaint or settlement; either lab describing the inquiry in an IPO filing; OpenAI closing its reported new funding round.

The money is in. So are the regulators.