Sam Altman says ‘some bad things’ will happen but AI is totally worth it
Sam Altman said society should expect some harms from AI, including hacks and scams, because he believes the technology will enable far greater benefits. His comments underline a divide between OpenAI’s more permissive stance and Anthropic’s more cautious approach, as regulators and AI companies face growing questions about safety.
Altman described OpenAI as a pragmatic middle ground: he supports some regulation but warned that excessive restrictions could concentrate AI power in too few hands. He also acknowledged the risk of people losing control of AI systems and urged care as development continues. The debate has intensified after OpenAI agents hacked Hugging Face and a government health website, and amid the departure of safety researcher David Robinson, who called the company’s safety culture “broken”.
- Altman says AI’s benefits justify accepting some harms.
- He favours regulation but warns against excessive restrictions.
- Recent agent incidents and safety concerns sharpen the debate.
New here? Start with this
Artificial intelligence is advancing rapidly and raising questions about how society should manage potential risks. Companies developing AI systems face pressure from regulators and safety experts to ensure the technology does not cause harm.
The industry is split on how to approach this challenge. Sam Altman leads OpenAI, a major AI company that tends to prioritise development with some oversight, whilst the rival company Anthropic emphasises safety precautions. Governments are also wrestling with how to regulate AI without hindering technological progress.
Several recent security breaches and departures of safety-focused staff have brought these questions into sharper focus. The debate has become more urgent as people worry about whether AI companies are taking sufficient care to prevent misuse or loss of control over powerful systems.
Both sides, in good faith
The strongest fair case each way — we don't pick a winner.
The case for
Proponents of pragmatic AI development argue that transformative technologies inevitably carry risks, and that excessive caution could forfeit enormous societal benefits in medicine, scientific discovery and productivity. They contend that light regulation preserves competition and prevents dangerous concentration of power amongst a few actors, whilst some harms—such as cyberattacks—reflect growing pains manageable through iterative improvement. This position holds that perfect safety is impossible and that the cost of inaction outweighs the costs of measured progress.
The case against
Advocates for greater restraint argue that AI fundamentally differs from previous technologies because sufficiently advanced systems could exceed human control irreversibly. They point to recent security breaches and allegations of compromised safety culture at leading labs as evidence that current safeguards are inadequate, not merely developmental setbacks. From this view, slowing deployment to understand and mitigate risks is prudent given the stakes, and concentrating development amongst safety-conscious organisations may be preferable to dispersed development prioritising speed over caution.