Sam Altman, chief executive of OpenAI, has said the world should be prepared to accept a degree of harm from artificial intelligence in return for the benefits the technology delivers. His remarks place the head of the company behind ChatGPT at the centre of a widening argument over how far governments and firms should go to restrain AI development.
Altman's position is that the potential gains — in productivity, scientific discovery and access to information — justify tolerating some negative outcomes rather than attempting to eliminate risk altogether. The intervention comes as policymakers in Britain, the European Union and the United States weigh new rules for advanced AI models, and as employers reassess how quickly automation might reshape their workforces.
The debate has been sharpened by evidence that AI is displacing more jobs than it creates. That finding sits awkwardly alongside the industry's argument that the technology will generate new categories of employment over time. For Altman, the trade-off is not costless, but he contends that slowing development carries its own price in foregone progress.
His comments also land amid persistent concern about safety failures, misuse and the concentration of AI capability in a small number of companies. Critics argue that accepting «some bad things» is easier for executives insulated from the consequences than for workers, consumers and communities exposed to them. Supporters counter that no transformative technology has arrived without disruption, and that over-regulation risks ceding leadership to less cautious competitors.
OpenAI has itself become a focal point for these tensions. The company has expanded rapidly while facing scrutiny over how its models are trained, deployed and governed. Altman has previously called for regulation of AI, even as his firm pushes to release more capable systems — a stance that has drawn accusations of talking tough while racing ahead.
For British business and political leaders, the question is practical as much as philosophical. The UK has positioned itself as a hub for AI investment while trying to build safeguards that do not deter firms from basing operations here. If the prevailing view among leading developers is that some harm is acceptable, regulators will face pressure to define exactly how much, and who bears it.
The jobs data complicates that calculation. If automation is already reducing employment faster than it creates roles, the case for a permissive approach rests on the promise of longer-term gains that may not arrive evenly. Regions and industries dependent on routine cognitive work could absorb the losses first, while the rewards accrue to shareholders and highly skilled workers.
Altman's framing is likely to intensify rather than settle the argument. It offers a clear position: progress should proceed, and society should absorb the costs. Whether voters, unions and legislators accept that bargain remains open — and the answer will shape not only AI policy but the wider relationship between technology companies and the public they serve.