Coinbase and Binance Reportedly Prepare Exchanges for Advanced AI Models
According to the new report, Coinbase, Binance and Fireblocks are seeking or testing Anthropic models as firms respond to AI-driven security threats and newly disclosed vulnerability findings.
Fact Check
The claim has three main parts: (1) Coinbase, Binance, and Fireblocks are seeking or testing Anthropic models; (2) this is connected to AI-driven security threats; and (3) there were newly disclosed vulnerability findings. The strongest directly aligned fetched source is the PANews article, which specifically attributes these details to The Information and names Coinbase, Binance, and Fireblocks with the described actions. The PANews article states Coinbase was in close communication with Anthropic on Mythos, Binance was testing vulnerabilities with Anthropic models, and Fireblocks had found a previously unidentified vulnerability using an earlier Anthropic model. That directly supports the user’s statement. The Information article URL was also observed via media tracing and search corroboration as the upstream report, but it could not be fetched in this run, so its validation remains null. Anthropic’s official pages, Project Glasswing and Assessing Claude Mythos Preview’s cybersecurity capabilities, do not mention Coinbase, Binance, or Fireblocks, but they strongly support the surrounding premise that Mythos is a restricted security-focused model and that Anthropic had recently disclosed major vulnerability-finding capabilities, including thousands of high-severity issues and a 27-year-old OpenBSD flaw. Together, these sources make the statement likely true, though confidence is medium rather than high because the core named-company reporting ultimately depends on a secondary fetched article plus an unfetched paywalled primary news report.
Summary
Coinbase, Binance and Fireblocks are seeking or testing Anthropic models to address AI-driven security threats, according to the new report. The update says Anthropic’s Mythos, launched in early April 2026 under Project Glasswing, identified thousands of zero-day vulnerabilities, including a 27-year-old flaw in OpenBSD. The new details expand the earlier report that major crypto firms were preparing their platforms for advanced AI systems by clarifying the cybersecurity rationale and naming Fireblocks alongside Coinbase and Binance.
Terms & Concepts
- Anthropic: An artificial intelligence company developing advanced AI models that businesses may use for tasks including security testing and automation.
- zero-day vulnerabilities: Software flaws that are unknown to the vendor or not yet patched, making them especially dangerous for security systems and infrastructure.
- OpenBSD: A security-focused open-source operating system; a reported 27-year-old flaw in it was among the vulnerabilities identified by Anthropic’s Mythos.