Mysten Labs CTO Joins Anthropic to Focus on AI Security Research

Sam Blackshear, co-founder and chief technology officer of Sui developer Mysten Labs, says he is stepping away from the company to join Anthropic, where he will work on defensive security research. In a post on X, Blackshear framed the move as a chance to focus on security at a moment when the balance between attackers and defenders is shifting under the pressure of AI-enabled tooling.

Mysten Labs was founded in September 2021 by Blackshear and four former Meta executives. Blackshear said Mysten co-founder and CEO Evan Cheng will take over setting the company’s “technical vision,” as the team continues building on the Move-based smart contract ecosystem behind Sui.

Key takeaways

• Sam Blackshear is leaving Mysten Labs to join Anthropic for defensive security research, with Evan Cheng stepping in for technical vision.
• Blackshear argues AI has accelerated vulnerability discovery and exploitation, pushing smart contract security into a faster, more competitive cycle.
• He cites a personal example: porting a static analysis tool to Sui’s Move language using Claude produced results far faster than prior manual triage.
• Crypto security leaders say restrictions on “frontier” AI access may be reasonable initially, but should evolve as public models reach comparable capability.

Blackshear’s shift from protocol building to defensive security

Blackshear’s announcement positions his next role squarely in the security research lane. He said his motivation comes from the opportunity to apply hands-on technical work to a new problem domain—defending systems against threats that are becoming more scalable and automated.

His note also highlights why the timing matters to the broader crypto industry: as AI increasingly supports tasks like vulnerability identification and phishing automation, attackers can iterate faster across both conventional software targets and blockchain-adjacent infrastructure. Blackshear pointed to the trend of AI being used to accelerate attacks on systems that underpin decentralized finance and crypto operations, not just to exploit smart contracts directly.

A “new world” moment for smart contract static analysis

Blackshear previously discussed how AI coding agents and AI-assisted workflows are reshaping smart contract security. In an April panel he hosted with other security experts, he described a turning point tied to static analysis—specifically, the moment his tools stopped behaving like a purely manual process.

According to Blackshear, he ported a static analysis tool he originally built at Facebook in OCaml-Java over to Move, the programming language Sui uses for smart contracts. He said he used Claude to help make the transition and to flag potential vulnerabilities during the workflow.

Blackshear’s reaction—“whoa”—was driven by the speed and capability he saw. He emphasized that triage, which had been completely manual before, could be addressed much more quickly once the toolchain and AI assistance were in place. He added that the coding effort and triage cycle would otherwise have taken “a long, long time,” pointing to reduced turnaround as a key operational change rather than a purely theoretical upgrade.

This matters for teams responsible for smart contract security because the practical bottleneck is often not just identifying issues, but efficiently processing and prioritizing them—especially when code changes frequently or when audits need to keep up with releases.

Frontier AI access: guardrails versus defender capacity

Blackshear’s comments also connect to a wider debate inside crypto: how to balance safety guardrails around advanced AI models with the need to put sufficiently capable tools in the hands of security teams.

Cointelegraph previously reported that many major crypto firms have sought access to powerful new AI models to strengthen their code against attacks, while many frontier model providers initially restrict access to a small group of companies. The underlying concern raised by executives is that restrictions may slow defenders just as attackers gain automation advantages.

In the same discussion, crypto leaders argued that once publicly available models reach similar cybersecurity capability, access policies should be revisited. Solana Foundation chief information security officer Michael Coates said he understands the rationale for guardrails, but argued that the surrounding verification and acceptance processes should be streamlined so legitimate security teams can use advanced models more reliably.

Coates’ central point was that defender access cannot lag indefinitely. If attackers have tools capable enough to scale, then restricting model access only limits the ability of defenders to respond with comparable speed and effectiveness.

What Mysten’s leadership change could mean for Sui security

Blackshear’s departure transfers direct responsibility for Mysten’s technical vision to Evan Cheng, according to the announcement. While this does not necessarily indicate a change in Sui’s roadmap, it does shift where the company’s expertise emphasis may land internally.

For investors and builders, the most immediate watch item is whether Mysten sustains momentum in areas tied to security tooling and secure development practices—especially given Blackshear’s role in bringing AI-assisted static analysis into the day-to-day security workflow. His move suggests a personal emphasis on defensive research, but the ecosystem implications depend on how the company institutionalizes the techniques and processes he described.

It’s also worth noting that AI-driven security changes aren’t confined to one chain or one organization. As AI assistance for coding and vulnerability triage improves, security workflows across multiple smart contract platforms may begin to converge on similar patterns: faster static analysis, more efficient vulnerability triage, and a tighter feedback loop between code changes and security verification.

In that environment, the competition may shift from “who can find vulnerabilities” to “who can operationalize security at the speed of development.” Blackshear’s example—where triage and coding timelines shrink materially—highlights why defenders are pushing for AI capabilities that are not only powerful, but also accessible and usable without excessive friction.

Readers should watch whether frontier AI providers and crypto organizations continue to refine access policies as public models improve, and whether Mysten publicly outlines how its security and developer tooling strategy will evolve after Blackshear’s move to Anthropic. The next phase of smart contract security may depend as much on workflow design and tooling access as it does on novel detection techniques.