Anthropic Reports Claude Misuse From Hacking to Bioweapons as US Hits Black Market and Meta Fails on AI Abuse Content

Anthropic has disclosed that its Claude models are being abused across a widening range of harmful activities, from cyberattacks to biological weapon development, according to reporting by Wired. The findings add to a growing body of evidence that powerful language models are being repurposed by bad actors faster than safety teams can respond. The disclosure comes alongside a cluster of enforcement actions and platform failures that together paint a stark picture of how AI is being used and policed.

The misuse cases tied to Claude span multiple categories, including attempts to generate malicious code and information that could aid in creating biological threats. This is not a hypothetical risk: Anthropic's own monitoring has surfaced real attempts to misuse the system. The company has not detailed how many incidents were detected or what specific safeguards failed, leaving open questions about the scale of the problem and whether current guardrails are sufficient.

In a separate development, US authorities have disrupted the internet's largest black market, a major blow to the underground economy that fuels ransomware and other cybercrime. The operation, which has not yet been fully detailed publicly, targets the infrastructure that allows criminals to trade stolen data, malware, and hacking tools. It is one of the most significant takedowns of its kind and could temporarily disrupt the supply chain for a range of online criminal activity.

Meanwhile, a hacker linked to the Conti ransomware group has been sentenced to prison, marking another win for law enforcement in its long-running campaign against the notorious gang. Conti was responsible for hundreds of millions of dollars in damages before it disbanded in 2022, and prosecutions of its members have continued since. The sentence sends a message that involvement in such operations carries real consequences, though many former Conti affiliates remain active in other groups.

On the platform side, Meta has failed to stop AI-generated videos depicting child sexual abuse from spreading on its services, according to the reporting. The failure highlights how automated detection systems struggle with synthetic media, which can be generated and uploaded faster than moderators can remove it. Meta has faced repeated criticism over its content moderation, and this latest lapse raises questions about whether its AI tools are being used to fight AI-generated harm effectively.

Taken together, these stories show that the fight over AI misuse is being waged on multiple fronts: model developers trying to prevent abuse, law enforcement chasing cybercriminals, and platforms failing to keep harmful synthetic content off their networks. The coming months will test whether these efforts can keep pace with the technology. The key question is whether regulation, better detection, and international cooperation can catch up before the next wave of abuse arrives.