Anthropic gaat AI-verzoeken voortaan transparant afwijzen om veiligheidsredenen

Anthropic will now inform users when their requests to its AI models are rejected or reduced due to national security restrictions, reversing its previous practice of silently downgrading certain queries. The company released Fable 5, a Mythos-class model, this week after initially withholding such advanced versions over concerns they could bypass cybersecurity defenses.