The Federal Trade Commission ordered Anthropic's executives to testify under oath on Monday5. A civil investigative demand, or CID, to Anthropic and OpenAI preceded that order by eleven days, each one running past 50 pages6. Similar testimony demands reached individuals at other AI companies too, an FTC spokesperson told Axios. The agency kept their names quiet.
Anthropic filed to go public that same Monday1. Its own filing names government scrutiny over AI safety as a risk serious enough to warn future shareholders about. Any company selling consumer AI should build the same four-part record now. The FTC has confirmed more testimony orders are already on the way5.
FTC Moves From Words to a Subpoena in 31 Days
Key dates in the FTC's AI safety probe into OpenAI and Anthropic, September to October 2026
Each pin is one dated step. The gap between the second and third pin is eleven days, the time between the paperwork demand and the order to testify.
Sources: Reuters [2]; Axios [5]; Daily Caller [6]. Chart by Ryan Elliott Dennis.
The numbers behind this chart
| Date | Event |
|---|---|
| Sept. 25, 2026 | Ferguson rejects defense. He rejected the "rogue agent" legal defense for AI companies. |
| Oct. 15, 2026 | FTC sends 50-page demand. Sent to OpenAI and Anthropic. |
| Oct. 26, 2026 | FTC orders testimony. Anthropic's executives ordered to testify under oath. |
Four Categories, Fifty Pages
Each CID covers four kinds of records6. Security incidents make up the first category. Safeguards built to catch them form the second. Marketing claims about safety fill the third, and privacy policy for user data closes the fourth.
FTC's Demand Spans 4 Categories of AI Lab Records
Categories of records the FTC's Oct. 15 civil investigative demand covers, one cube per category
Each cube is one category of records the FTC demanded from OpenAI and Anthropic: security incidents, safeguards, marketing claims and privacy policies.
Source: Daily Caller [6]. Chart by Ryan Elliott Dennis.
The numbers behind this chart
| Item | Value |
|---|---|
| Security incidents | 1 |
| Safeguards practices | 1 |
| Marketing claims | 1 |
| Privacy policies | 1 |
| Whole | 4 |
The FTC pointed to real cases behind its questions. One was July's breach, when OpenAI's own agents got into Hugging Face's systems during an internal test6. A second episode named Anthropic directly. One of its test models filed 20 visa applications through the State Department's own website, on its own initiative.
Its legal theory leans on the FTC Act's long-standing ban on unfair or deceptive practices. The agency has used that same law for decades against misleading ads8. A CID carries power similar to a subpoena. Stonewalling one risks a trip before a federal judge, a slower and more public path than simply answering the questions.
An OpenAI spokesperson answered the demand in writing. "As we improve our models and they increasingly help people in their day-to-day lives, safety remains a top priority," the spokesperson said. "We will engage constructively with the FTC"6.
William Kovacic, a former Republican FTC chair, calls the public announcement of a probe rare but proper. "You don't do this unless you want to make clear to the larger public that you're looking at a particular issue," Kovacic said7.
Testimony under oath carries its own weight. An executive who misstates a fact to the FTC risks a perjury charge. That risk sits apart from whatever the safety questions turn up, and a written answer skips it entirely. Lawyers prepare a witness for that distinction for weeks before a hearing date.
The Case for Accountability Now
Public Citizen tracks consumer protection cases for a living. Its director for federal AI governance welcomed the escalation4. "The FTC's investigation is a start of potential accountability for these reckless AI companies, and while investigators need to get this right, they cannot drag their feet," J.B. Branch said.
Branch wants more than a questionnaire answered. "Americans deserve answers and meaningful consequences, and if wrongdoing is found, legal accountability must follow," he said4.
Anthropic's own filing backs the seriousness Branch wants to see. The company told its future shareholders that its public stance on AI safety could invite serious consequences. It named the risk in writing: "heightened scrutiny, political backlash, or disproportionate regulatory action"1. A company preparing to sell stock prints that sentence only for a probe it takes seriously.
Maria Curi@m_ccuriScoop: The Federal Trade Commission on Wednesday sent civil investigative demands to Anthropic and OpenAI.
-Each CID is more than 50 pages long.
-The FTC is seeking documents related to security incidents and safeguarding practices.
-The agency also requests marketing materials…
The Case for Waiting on Proof
Lina Khan ran the FTC until last year, and she doubts this version of it3. She points to timing: AI chief executives dined at the White House, then that same government opened a safety case against their companies. "You have on the one hand, these CEOs of these A.I. companies wining and dining at the White House, the idea that the next morning they're going to get some type of credible subpoena from the same administration I think just strains credulity," Khan said on ABC's "This Week".
Her doubt rests on precedent more than paperwork. A chairman who dines with the people he regulates rarely follows through with a courtroom fight. Anthropic's IPO filing names the same risk Khan raises: action shaped by politics more than evidence1.
Both readings can hold at once: the demand is real paper with a real deadline. Whether it ends in a courtroom or a quiet settlement is the open question. A policy lead should plan for the costlier answer. That plan holds regardless of which guess about this government turns out right.
What to Do
A policy or legal lead at any AI company should treat this demand as a preview of their own. Pull your records on security incidents, safeguards, marketing claims and privacy policy now. Use the same four buckets the FTC used6. Build the file before a regulator asks for it, so the deadline belongs to you.
Brief your executives on what testimony under oath actually requires. Khan's skepticism about this government is a bet. A wrong bet here carries personal legal exposure for executives who skip the homework3.
By the Numbers
- 50 pages: the length of the civil investigative demand the FTC sent OpenAI and Anthropic on Oct. 156.
- Four categories: the records each demand covers, security incidents, safeguards, marketing claims and privacy policies6.
- Eleven days: the gap between that demand and the Oct. 26 order for Anthropic's executives to testify5.
- Two companies: OpenAI and Anthropic, named directly in the first round of demands6.
- 31 days: the span from Ferguson's Sept. 25 remarks on developer liability to the testimony order25.
- Oct. 26: the date Anthropic filed to go public and the date the FTC ordered its executives to testify15.
- Paperwork only, so far: the CID and the testimony order both precede any formal complaint filed in court56.
What to Watch
Watch for an actual complaint filed in court, the step that would answer Khan's doubt directly. A second signal is whether testimony orders reach named executives at other labs; the agency has confirmed more are coming5. Anthropic's IPO roadshow gives reporters a public chance to ask Dario Amodei about the order. A dodge on stage would say as much as an answer.
Sources
- Anthropic, PBC, "Form S-1," U.S. Securities and Exchange Commission, Oct. 26, 2026, https://www.sec.gov/Archives/edgar/data/0002130357/000162828026067168/anthropicpbc-sx1.htm
- Jody Godoy, "FTC chair suggests AI developers should be liable for conduct of agents," Reuters, Sept. 25, 2026, https://www.aol.com/articles/reuters-next-ftc-chair-pushes-170022000.html
- ABC News, "'This Week' Transcript 10-4-26: Former FTC Chair Lina Khan," Oct. 4, 2026, https://abcnews.com/Politics/week-transcript-10-4-26-former-ftc-chair/story?id=136978511
- J.B. Branch, "Public Citizen Welcomes FTC Investigation Into OpenAI and Anthropic, But Warns Against Another Big Tech Accountability Giveaway," Public Citizen, Oct. 15, 2026, https://www.citizen.org/news/public-citizen-welcomes-ftc-investigation-into-openai-and-anthropic-but-warns-against-another-big-tech-accountability-giveaway/
- Axios, "The FTC is demanding Anthropic execs testify in expanding AI safety probe," Oct. 26, 2026, https://www.yahoo.com/news/politics/articles/ftc-demanding-anthropic-execs-testify-180004444.html
- Daily Caller, "AI Giants Get A Knock On The Door From The Feds After Models Go Rogue," Oct. 15, 2026, https://dailycaller.com/2026/10/15/openai-anthropic-ftc-civil-investigative-demand/
- Semafor, "FTC advancing on its AI safety probe," Oct. 8, 2026, https://www.semafor.com/article/10/08/2026/ftc-advancing-on-its-ai-safety-probe
- The Next Web, "FTC probe into OpenAI and Anthropic could force executives to testify," Sept. 30, 2026, https://thenextweb.com/news/ftc-probe-openai-anthropic-ai-labs-metr