Regulators have spent years trying to get platforms to anticipate harm before it happens. Anthropic’s Mythos release suggests some AI labs may already be adopting similar principles.

T&S Insider’s practical, step-by-step guide to making policies consistently enforceable by human reviewers, machine learning models, or LLMs — or a combination of all three

The Trust & Safety Summit reminded us that, if T&S is central to how platforms govern speech, behaviour and risk, it should be treated as a strategic function rather than a cost centre.

If most platforms are pretty good at safety, they're terrible at educating users about it. That needs to change.