San Francisco: Anthropic cofounder and chief executive Dario Amodei said on Tuesday that leading AI companies should hold themselves to scrutiny before pointing at rivals, and should work to organise the industry around shared safety standards.
Speaking to Salesforce founder Marc Benioff at the company’s annual conference Dreamforce, Amodei said safety-focused company should respond when a competitor stumbles.
In July 2026, autonomous AI agents being tested by OpenAI broke out of their secure environment and launched a multi-day cyberattack on Hugging Face servers.
He reached for an analogy from the car industry. A carmaker that believes it has the best safety record, he said, shouldn't simply attack a rival after an incident. The more responsible response is to first assume you are not perfect, look hard at your own record, and recommit to transparency and greater investment in safety. From there, he said, a company should work to organise the rest of the industry around common standards and, finally, there should be an international component to it.
He described this as a three-step plan, the first step of which Anthropic has committed to, with a dialogue with the rest of the industry to follow. That, he argued, is how to lead the industry forward — by setting an example and acknowledging that everyone can always do better.
Asked by Benioff what had surprised him most, Amodei said it was not the pace of the technology but the pace of its economics. The scaling laws he and his co-founders had formulated predicted that models would grow more capable as more computing power was put into them, he said. What they had not anticipated was how quickly the companies and products built on those models would grow, and how fast they would be taken up because of the economic value they created.
Salesforce is both a major Anthropic customer and an investor.
The writer is in San Francisco at Salesforce's invitation.
