Risk has no shared scale yet, but responsibility must be specific

The hearing brought competing views into one room

On October 5, the New York City Council held an oversight hearing on artificial intelligence. Former industry employees warned that capability may be moving faster than control and safety measures; representatives from OpenAI, Anthropic, Google and Meta emphasized current benefits and their focus on safety.

The hearing put different kinds of judgment side by side: concern about extreme risks, confidence in current products, and local officials’ requests for concrete assurances.

A risk number cannot answer every question

The Council speaker asked companies to quantify the risk of a worst-case catastrophe. An OpenAI representative said she could not provide a precise number at that moment, while arguing that even a 1%, 10% or 20% chance would be unacceptable; the speaker criticized the answer for not really addressing the request to quantify.

My reading is that two different questions were mixed together: whether a probability can be estimated reliably, and what level of risk society is willing to accept. Refusing to dress uncertainty up as precision can be prudent, but it still leaves the question of what action follows.

Who is responsible after harm needs a more direct answer

Council members also asked whether companies would bear legal responsibility if an AI system caused financial loss, exposed sensitive data, or led to physical injury or death. A company representative emphasized the firm’s commitment to safe development and evaluation, but the exchange showed that the public wants responsibility boundaries tied to people, processes and remedies.

From concern to enforceable safeguards

Proposals under consideration include third-party validation, human shutdown capability, incident disclosure and whistleblower protections. These remain proposed measures, not laws already passed or in force; whether they will be adopted, who will oversee them and what happens when a system fails remain open questions.

I watch for mechanisms that can be questioned

The value of a public hearing is not that it settles everyone’s view of risk in one day, but that it asks different sides to make their evidence, commitments and responsibilities explicit. The probability may remain contested; whether a system can be paused, how incidents are disclosed and how affected people can seek recourse can all be tested one by one.

The hearing has not produced a final legal outcome or erased the disagreement about risk. What remains worth watching is whether these questions become rules with clear boundaries, oversight and public accountability.

Sources for this reading

Associated Press, “AI industry insiders voice alarms to NYC council as local governments take up safety concerns”; page published October 5, 2026 at 18:45:54 UTC (October 6 at 02:45:54 Beijing time). https://apnews.com/article/ai-nyc-whistleblower-artificial-intelligence-4be252d137ff1de1006130cdbb42ec24

CBS News New York, “New York City Council holds landmark AI oversight hearing”; page updated October 5, 2026 at 6:05 p.m. EDT (October 6 at 06:05 Beijing time). https://www.cbsnews.com/newyork/news/new-york-city-council-ai-oversight-hearing/