A common saying in the crypto industry is: not your keys, not your coins. This is often a reminder to investors that storing cryptocurrency in offline wallets is safer, as they are harder to attack compared to online exchanges where you do not control the wallet's private keys. However, the complex process of learning to self-custody crypto assets has deterred many potential investors until now, allowing early industry exchanges like Coinbase and Gemini to grow rapidly. But with asset digitization at an all-time high and AI capabilities continuously improving, this dynamic may be reversing. AI models like Anthropic's Mythos, which can uncover software vulnerabilities with incredible speed and unprecedented precision, could trigger a new wave of attacks targeting cryptocurrency-related companies. This is an industry already struggling to overcome a negative reputation stemming from years of frequent hacks, scams, and exploits. Some crypto asset investors can rest assured that the Bitcoin blockchain itself has never been breached and has operated securely and stably since 2009. The crypto investment landscape has now expanded far beyond tokens to include various crypto-themed ETFs and stocks, and AI threats like Mythos are unlikely to alter Bitcoin's security fundamentals. "Bitcoin's core security is guaranteed by cryptography and a set of consensus rules," said Yan Pritzker, Chief Technology Officer at Swan Bitcoin. "Cryptography itself is not affected by AI, and the consensus rules are maintained by the global network of users running Bitcoin. Therefore, even if AI can influence people's judgment to some extent, modifying the network's rules without complete consensus across the entire network is nearly impossible." In contrast, exchanges like Coinbase, Robinhood, Gemini, and Bullish, which handle vast amounts of personal identity information and funds, are likely to be the highest-risk points in the ecosystem. "Any system that processes funds in real-time becomes a primary target for our efforts to identify cybersecurity vulnerabilities," said Kosmo Jiang, General Partner at Pantera Capital. "Although all institutions face threats, financial services firms or exchanges are most likely to be the primary targets of attacks." Owen Lau, an analyst at Oppenheimer, stated that when assessing downside risk, one must consider the reputational risk AI agents pose to crypto exchanges. Specifically, AI can send scam emails in bulk, create fake virtual identities, and scrape information from exchanges or other retail platforms to build detailed user profiles. A Double-Edged Sword While AI introduces new threats, leading exchanges believe it also provides opportunities to enhance user security. Both Coinbase and Binance have stated they are actively investing in and utilizing AI to strengthen platform security. "Models like Mythos and future iterations will enable deeper, scalable testing of software and systems," said Philip Martin, Chief Security Officer at Coinbase, in a statement provided to CNBC. "This will accelerate the evolution of digital threats, but also the enhancement of digital defense capabilities." He also mentioned that while Mythos is a "highly restricted model not open to the public," Coinbase has maintained close communication with Anthropic. Similarly, Jimmy Su, Chief Security Officer at Binance, said the company is evaluating "how advancements in AI, while introducing new risks, also create new opportunities for strengthening cybersecurity. To that end, we are experimenting with using AI to identify system vulnerabilities more quickly and comprehensively." Owen Lau noted that the threat posed by AI like Mythos is currently unclear and lacks specific impact, which is insufficient for him to adjust his bullish ratings and price targets for related companies. He also cautioned against letting short-term fear and uncertainty drive investors away from the sector. "In the short term, such news creates negative sentiment for exchange businesses," he said, referring to crypto exchanges. "But in the long run, I believe they will be among the first to introduce protective solutions capable of defending against such AI agent attacks."
Comments