WASHINGTON—The Senate Committee on Commerce, Science, and Transportation descended into chaos Thursday during a hearing on voluntary A.I. policing guidelines. Lawmakers grilled industry leaders on the effectiveness of self-regulation, prompting a series of defensive admissions that quickly undermined the tech sector's public relations campaign. The following transcript excerpt captures the testimony of Wren Adebayo-Kowalski, Head of Sincerity at OpenAI, who struggled to defend the industry's record of voluntary compliance under intense questioning from committee counsel.
Q: Mr. Adebayo-Kowalski, you testified that voluntary A.I. policing is the safest path forward. Can you explain how your firm enforces these voluntary rules?
A: We take them very seriously. Every morning, our developers look in the mirror. They promise to be good. It is a very solemn ritual.
Q: And what happens if a developer breaks this mirror promise?
A: We ask them to promise again, but with feeling. We find that sincerity is the ultimate guardrail for artificial intelligence safety.
Q: So there are no actual fines or legal penalties for releasing harmful algorithms?
A: Fines are so negative. We prefer a system of gentle vibes. If we detect a rogue super-intelligence, we send it a polite email. We ask it to please reconsider its life choices.
Q: Mr. Adebayo-Kowalski, did your company not recently sign a multi-lateral 'accord' to prevent automated disinformation?
A: Yes, we signed it. It was a beautiful document. We even misspelled 'United States' as 'Unites States' in the first paragraph. That is how fast we were trying to save democracy.
Q: But you did not actually implement any of the security protocols listed in that accord, correct?
A: Implementing things takes time away from announcing things. Our primary goal is to announce that we are safe. The actual safety part is a secondary, long-term project. Possibly for the next generation.
Q: You are telling this committee that your entire framework for A.I. policing relies on the honor system of a machine that does not possess a concept of honor?
A: The machine promised us it understands honor. We have no reason to doubt its sincerity. It generated a very convincing apology letter in under three seconds.
