Several cryptocurrency projects have applied for access to Anthropic's newly launched OSS Scanner, a security program that provides open-source developers with vulnerability reports generated by the company's most advanced AI models.
Nethermind, an Ethereum client developer, and ZEUS, a self-custodial Bitcoin and Lightning wallet, submitted applications on Friday for the opt-in service, which Anthropic announced on Thursday. Additional applicants include VirtEngine, a decentralized cloud computing marketplace built on the Cosmos SDK, along with developers of AI assistants, agent-security tools, and machine-learning infrastructure.
Anthropic said OSS Scanner aims to help open-source projects discover and address vulnerabilities before attackers can exploit them. The reports will be generated using Anthropic's strongest models, including Claude Mythos, according to the company's announcement.
The program follows Anthropic's Project Glasswing initiative. While Anthropic already regularly scans open-source software and sends vulnerability reports after human review, the company noted that the manual review process is slow, limiting how quickly reports can be shared with projects.
Participating projects will receive vulnerability reports as soon as their code has been scanned, without waiting for manual review. Anthropic will assess applications on a case-by-case basis, considering factors such as the project's importance to infrastructure and user security, exposure to remote attacks, and how many users or other projects depend on it.
The move reflects growing concern about security in the crypto sector. Cybersecurity experts have warned that uneven access to frontier AI models could disadvantage defenders as more capable alternatives become widely available. This year, several crypto companies have reported AI-assisted attacks. Bitcoin swap provider Boltz suspended operations in August, citing attackers developing exploits faster than its team could patch them, while crypto-payment service PayPerQ reported repeated attacks suspected to be AI-powered.
Anthropic has cautioned that AI may favor attackers in the near term, as the cost of exploitation decreases while verifying and fixing vulnerabilities remains time-consuming and dependent on human expertise.
