… While Garraghan says other chatbots are equally vulnerable to the kind of social attack the researchers used on Claude, they focused on Anthropic given the company’s self-proclaimed attention to safety and strong performance in other red-teaming efforts, including a study testing whether chatbots w… …
… It still limits: Fable deferred to Opus when I asked it about sarin gas, a highly toxic nerve agent. …
… And the same skills that can break a chatbot could soon be used to break the AI agents coexisting with us in the real world — booking meetings, managing calendars, ordering food, handling customer service — and safety teams will need to ensure models respond appropriately to very different kinds of… …
… We were founded to do the safety science research, the standards development, and the public advocacy. …
… Image: https://platform.theverge.com/wp-content/uploads/sites/2/2026/05/yarbo-email-andreas-0.jpg?quality=90&strip=all Image: https://platform.theverge.com/wp-content/uploads/sites/2/2026/05/yarbo-email-andreas-2.jpg?quality=90&strip=all That’s why Makris decided to do something that security resea… …
… I/ think a lot of the underlying safety researchers would say potentially risking breaking the country, breaking the world, and breaking millions of people whose jobs and safety hang in the balance — that’s what’s unique about it. …
… The source familiar with the negotiations said that the Amazon research was explicitly mentioned in conversations with the US government. …
… That’s a policy that’s in place, and there are agents that are following those policies, et cetera. Human agents you mean. Your current agents. Human agents. …
… We work with public safety. …
… The product is trust and safety. It is content policy, which I think Elon learned the hard way. Now, we see it. SpaceX filed their S-1, and we got to see their revenue cut in half since the purchase. I think he learned very clearly that content policy, trust and safety, is the product. …