The Pope, Bernie Sanders and Steve Bannon just handed Brussels a mandate to rein in AI

Here is a legislator’s dream: an industry fights against regulation, brands regulators as ignorant bureaucrats who stifle growth and spends millions in lobbying to carry these points home. And then it changes its tune: it admits that their product is extremely dangerous, promises to become more cautious and urges more regulation. This dream came true last weekend. Anthropic’s CEO, Dario Amodei, published an essay calling for embedded independent evaluators inside AI companies, coordination among frontier labs in democratic countries, and eventually an international agreement that includes China (in fairness, and in contrast to others, he asked for regulation for a long time). Anthropic is committing unilaterally to the first step: giving independent evaluators permanent, employee-level access to verify its safety practices, with the right to publish their findings without the company’s editorial control. Sam Altman agreed and also promised independent evaluations. Elon Musk agreed with the overall point. There are two schools of thought on what this means. The cynical view The cynical school says that the big AI companies realise that they will not recoup their massive investments any time soon. It was for this reason that analysts already expected OpenAI’s delay of its stock market debut. It is a good line to take. Instead of saying: “this technology may be less groundbreaking and profitable to businesses than we thought“, they double down on the groundbreaking. They are saying: “This is so groundbreaking, we need to slow down to keep it safe." There are variants of the cynical school.