I've been getting questions about the AI warnings in the news. Is this real, or is the coverage getting carried away? There are real reasons to be concerned, and I think we need to understand them without treating every prediction as something that's already happened.
In this week's essay, I look at the calls to slow frontier AI development, including Bill Gates' perspective and Dario Amodei's proposal to pace progress. Independent testing and better oversight could help. Getting competing companies and countries to accept sustained limits is a much harder proposition.
The companies have invested heavily and face pressure to deliver returns. Governments see AI as a source of economic and strategic power. And any agreement would need credible ways to check whether the other side was keeping its promises. Those pressures reinforce each other, which is why I think a coordinated slowdown is unlikely.
There are difficult choices within the safeguards too. Keeping the strongest systems under controlled access might help prevent misuse. It could also leave a small group of unelected companies with extraordinary influence over the rest of us.
That's why I want a much wider contribution to this conversation. Using AI on a task you understand, and checking its work, can help you see what is changing. Then consider what happens when that capability becomes cheap and widely available. You deserve a say without having technical credentials. Take those questions into your workplace, your community, or a conversation with your MP, and ask how you'll be represented.
Sources & citations
- Scaling laws research
- Bill Gates: The turbulent AI era is here
- Anthropic: September threat intelligence report
- OpenAI: An Alien Mind
- Business Standard: Anthropic researcher’s departure
- Dario Amodei: We Must Pace the Frontier
- Microsoft: FY2026 Q4 earnings call
- IMF: Technology investment and growth
- Reuters: Trump’s 14 September response
- China’s official response
- AP: Diplomatic context
