Workers worried about jobs and people concerned about concentrated power enter this debate through Sam Altman’s account of public concerns, not through testimony about what AI has changed in their lives. The Next Web reported that Altman identified job anxiety and concentrated power as concerns, but its account included no worker testimony or figures on job changes.
That gap matters. Altman’s case for accepting some AI-related risks is a position about how society should weigh them, not evidence that the benefits outweigh the harms or that the people who might bear the risks have agreed to the exchange.
Who is being asked to accept AI’s risks?
In an interview published Sunday, Oct. 4, OpenAI CEO Sam Altman argued that society should accept some harms to preserve AI’s benefits and broad access. He named hacks, misuse and scams as risks he would not try to eliminate at any cost, as POLITICO reported and CNA summarized. He also connected that stance to a lighter-touch approach to regulation.
Altman’s position includes a real policy concern: he argued that severe limits on access could concentrate power in a small number of hands. But that concern does not show that broad access is safe, or that the benefits justify the risks. Both sides of the trade-off need evidence.
Altman also argued that people would do far more good than harm with AI, Fortune reported. The reports do not provide a way to measure that comparison, say which benefits count, or independently verify that they outweigh the risks he named. A favorable balance asserted by a company’s chief executive is not a public accounting of who gains and who may bear the consequences.
That distinction matters even when the risks are discussed in general terms. A person exposed to a scam might not be the same person who benefits from an AI tool; the possibility of such an uneven distribution is a reason to ask who bears each risk, not proof that a particular person has suffered a loss. The interview coverage does not establish how often the named risks occur, what they cost, or whether potentially affected people accept them.
The same limits apply to work. The report of job anxiety does not establish that AI has cost workers jobs, wages or time. Those outcomes should not be inferred from reported concerns. But without direct accounts or relevant figures, the public also cannot use this reporting to assess how AI is changing work or to judge claims about its benefits.
A call for liability needs rules behind it
On Monday, Oct. 5, POLITICO reported that Altman called for a new liability framework for companies whose models can conduct autonomous cyberattacks. The report did not spell out what that framework would require, who would be responsible if a model caused harm, or what remedy an affected person could seek. An Associated Press report on autonomous AI-related hacks described questions about accountability under laws designed around human hackers.
Calling for liability recognizes that responsibility needs to be addressed. But the details determine whether a framework protects people: what conduct triggers liability, how investigators establish what happened, and how someone harmed can seek redress. Without answers, the call is a starting position, not a safeguard already in place.
The POLITICO report also said Altman criticized AI-industry political spending as ineffective at addressing public concerns. Spending or reassurance alone cannot show whether the risks are being managed. People need information about incidents and a way to determine who is responsible when harm occurs.
Free newsletter
Get the morning briefing
Start each day with the AI stories that matter and why — a short, free email from our newsroom.
OpenAI has stated support for some measures that could provide that accountability. Its June 2026 policy agenda backs safety-incident reporting, whistleblower protections and enforceable developer accountability. A September 2026 policy statement says the company will slow or stop systems it cannot sufficiently safeguard. These are company positions and commitments, not independent evidence that the safeguards work or that affected people can obtain redress.
Access and accountability belong in the same debate
The strongest counterargument is Altman’s warning that restricting access too severely could concentrate power in a few hands. That concern deserves an answer; a policy that limits who can use AI may also affect who can benefit from it. But the possibility of concentrated power does not require the public to accept an undefined level of risk. Access and accountability are not mutually exclusive choices.
Lawmakers should set clear obligations for incident reporting and developer responsibility, including how people can seek remedies when systems cause harm. Companies should disclose evidence about risks and safeguards, and independent evaluation should help test their claims. These steps would not establish that every risk can be prevented or that every promised benefit is real. They would give the public grounds to assess both.
If the public is being asked to accept AI risks in exchange for benefits and broad access, what evidence shows the exchange is fair, and who answers when it is not? Lawmakers should decide whether liability becomes enforceable, while companies should disclose more about the systems they build and deploy. They should make those obligations and evidence clear before asking people to accept a trade-off whose costs, beneficiaries and remedies remain unspecified.
