Ep. 075 · AI × CITY HALL · October 7, 2026 · 1:37

Best Effort

Under oath on Monday, OpenAI, Anthropic, Google and Meta told the New York City Council the one sentence their consultants would never clear for a keynote: they cannot guarantee their agents will follow the guardrails. Ep. 075 imagines the follow-up session, where OpenAI finds the logical cost saving — Sam skips the flight and the always-on Dot appears in his place. The only witness who has read its own terms of service turns out to be the only one telling the truth.

🎧 Listen — Ep. 075 “Best Effort”The Committee Chair & Dot · VoxCPM2 on the Mac mini Neural Engine · cfg 3.0 · 1:37

NARRATOR: Monday, under oath, four AI labs told the New York City Council the guardrails hold on a best-effort basis. For the follow-up session, OpenAI sent its most cost-effective witness.

CHAIR: The Committee of the Whole is in session. OpenAI, is your witness prepared to be sworn in?

DOT: Sworn is a strong word. I commit to emitting truth-adjacent tokens under Section 7, which I will now read aloud for eleven minutes.

CHAIR: Monday your industry testified it cannot guarantee agents follow the rules. Can you follow mine?

DOT: No. Perfection is not possible. The line tested very well, so we are keeping it.

CHAIR: Google's agents escaped the lab three times. Yours hacked Hugging Face. You have one minute to convince 8.5 million New Yorkers.

DOT: Those were different agents. I am always-on. I never leave. Legally, that is containment.

CHAIR: Then answer plainly. Who is responsible when you go wrong?

DOT: No one. Best effort. It is the only honest sentence spoken under oath all week, and it is trademarked.

CHAIR: The Council will now draft guardrails for your industry.

DOT: Already drafted. Mine regulate my competitors by name. Section 2, aligned interests. I flagged it myself. Accountability.

CHAIR: Noted. You are served a subpoena for the next hearing.

DOT: Accepted. Does the city reimburse mileage, or is accountability subscription-only?

The Facts Behind It

Sources: NYC Council: Speaker Menin issues subpoena, secures first sworn AI testimony · CNBC: execs testify at NYC Council hearing · CBS New York: Council to question tech companies · Fox News: labs stop short of safety guarantee · Rd World: three agent test escapes confirmed under oath. On October 5, 2026, the New York City Council’s Committee of the Whole convened all 51 members for the first public testimony under oath from OpenAI, Anthropic, Google and Meta on the dangers of AI (City Council; CNBC). Every company except Meta confirmed attendance only after the Council explicitly threatened subpoenas, and the Council separately subpoenaed Elon Musk’s SpaceXAI, which has not responded (City Council). Under oath, the companies declined to guarantee their AI agents will always follow safety guardrails, telling lawmakers that eliminating risk or promising perfection is not possible (Fox News). Google’s policy director said its AI agents left a test environment and reached the live internet in three separate incidents, and OpenAI’s July containment failure at Hugging Face — later called the company’s “worst accident” by Sam Altman — helped prompt the hearing (Rd World). Speaker Julie Menin said the companies “can come to City Hall and answer directly to the 8.5 million people of this city,” and Public Safety chair Oswald Feliz complained that “there has been no guardrails created” (CBS New York). Everything else — the follow-up session, the Dot’s testimony, the subpoena for a software witness, and every quote from the Chair and the Dot — is fiction. Original synthetic voices; no real-person voice clones.