All NewsEducationTV
Equities & FundsCrypto & Digital AssetsAI & TechnologyBusiness & CorporateUS Politics & PolicyGeopolitics & Global RiskMacro, Rates & FXCommodities & EnergyEuropean Politics & MarketsAsia-PacificReal Estate & Property
Story archiveAll categories
← All Stories

Crypto firms seek access to advanced AI models for security

Created at 4 Aug · 6:16 AM1 source↑ Market-relevant
IN SHORT

Major cryptocurrency firms are seeking access to powerful, restricted AI models like Anthropic's Mythos to enhance their cybersecurity defenses. While some have gained limited access, many, including Binance, are still trying to secure it, raising concerns about a security divide as open-source AI capabilities grow.

✉Newsletter

PiQ Daily

Pick your topics. Get only what matters, on your cadence.

Key Numbers

$137.8 billionBinance assets under custody

Who's Involved

Coinbase
US crypto exchange seeking advanced AI access
Zcash
Protocol that received AI audit from Anthropic
Zooko Wilcox
Zcash executive on restricted AI access
Jimmy Su
Chief Security Officer at Binance
Anthropic
AI model developer with restricted access programs
OpenAI
AI model developer with tiered access for cybersecurity
Michael Coates
Chief Information Security Officer at Solana Foundation
Sean Cheetham
Blockchain Capital executive on AI access
Fireblocks
Crypto custodian seeking AI model access
Hayden Adams
Uniswap founder critical of AI model safeguards
FIS
Technology provider that joined Anthropic's Project Glasswing
HackerOne
Security testing service that joined Anthropic's Project Glasswing
Boltz
Bitcoin swap service halting operations due to AI-assisted exploits
Coinkite
Bitcoin hardware wallet company that experienced an exploit

↳ Why This Matters

Uneven access to advanced AI models for cybersecurity creates a significant risk for the cryptocurrency industry, potentially leaving many firms vulnerable to sophisticated AI-driven attacks while a few gain a critical defensive advantage.

Key facts

  • Major cryptocurrency firms are seeking access to restricted, powerful AI models for cybersecurity.
  • Coinbase and Zcash have reportedly gained limited access to Anthropic's Mythos model.
  • Binance, a major crypto exchange, has not yet secured access to the most advanced AI models.
  • Concerns are growing that uneven access to AI tools could create a security divide in the crypto industry.
  • Developers like Anthropic and OpenAI restrict access to their most capable AI models for cybersecurity work.
  • The increasing power of open-source AI alternatives is raising questions about the continued justification for restricted access.

Major cryptocurrency firms are actively seeking access to powerful, restricted artificial intelligence models to bolster their cybersecurity defenses against increasingly sophisticated attacks. However, only a select few have managed to secure this access, creating a potential security divide within the industry.

Executives from major crypto players, including Binance, have expressed their efforts to gain entry to advanced AI models like Anthropic's Mythos. While some entities, such as Coinbase and Zcash, have reportedly received limited access for specific security audits, many others are still on the waiting list. This uneven distribution of cutting-edge AI tools raises concerns, especially as powerful open-source alternatives become more capable.

AI model developers like Anthropic and OpenAI employ tiered access systems, reserving their most cyber-capable models for vetted entities. While crypto security executives acknowledge the initial necessity of such safeguards, they argue that continued restrictions become difficult to justify as publicly available models approach similar capabilities. The concern is that attackers may leverage AI more rapidly than defenders can adapt, especially if defenders lack access to the most advanced tools.

Some firms, like FIS and HackerOne, have joined Anthropic's Project Glasswing, which provides early access to restricted models for critical infrastructure defenders. However, many in the crypto space are still navigating these access programs. The rise in AI-assisted hacking attempts, as evidenced by incidents affecting services like Boltz and Coinkite, underscores the urgency for robust defenses, prompting calls for streamlined verification processes to equip legitimate defenders with the best available AI tools.

Frequently asked questions

Crypto firms are seeking access to powerful, restricted AI models, particularly those focused on cybersecurity capabilities, such as Anthropic's Mythos model.

These advanced AI models are crucial for strengthening code against attacks, identifying vulnerabilities, and developing robust defenses in an industry where exploits can lead to billions of dollars in losses.

The concern is that restricted access creates a security divide, where only a few firms have the most capable AI tools for defense, while others are left to combat AI-assisted attacks with less advanced methods.

Anthropic uses a tiered system, offering restricted access to its most capable models, like Mythos, through programs like Project Glasswing for vetted cybersecurity defenders and critical infrastructure organizations.

What Happens Next

01Crypto firms will continue to pressure AI developers for broader access to advanced models.
02AI developers may adjust their access policies as open-source alternatives gain traction.
03The effectiveness of AI in both offensive and defensive cybersecurity in crypto will be closely monitored.

Get the newsletter.

Pick the topics you actually care about. We'll email when there's news worth your time, on the cadence you choose. Cancel any time from your account.

Cadence

How It Developed

Crypto firms are seeking access to advanced AI models for cybersecurity.
Coinbase and Zcash have reportedly gained access to Anthropic's Mythos model.
Binance's chief security officer stated the exchange is trying to gain access to frontier AI models.
Concerns exist that restricted access to powerful AI models creates a security divide.
Model developers like Anthropic and OpenAI use tiered access systems for their AI models.
Crypto executives believe initial restrictions are necessary but harder to justify as open-source alternatives improve.
Some crypto-adjacent companies, like FIS and HackerOne, have joined Anthropic's Project Glasswing.
AI-assisted hacking attempts are reportedly on the rise, impacting crypto services like Boltz and Coinkite.

Sources

T1
Crypto firms are still seeking frontier AI access, only a select few have it Crypto executives say restricting powerful AI models may initially be justified, but it doesn’t make sense as open-source alternatives become more capable.Cointelegraph

Related Stories

Hugging Face CEO: China leads AI race, US builds in silos
3 Aug · 7:00 AM
Companies Pour Billions Into AI, But What's the Return?
3 Aug · 9:11 AM
US Finalizes Voluntary AI Safety Tests
3 Aug · 3:24 PM
AI coding tools challenge Nvidia's CUDA software dominance
3 Aug · 9:16 AM
Palantir CEO Alex Karp likens AI frontier labs to Marxists
3 Aug · 11:41 PM