AI Is Helping Identify Tech Threats—And Zcash Is The Latest Example

The latest forms of frontier AI aren’t limited to interacting with users, creating images, or writing code. Researchers are increasingly using systems such as Anthropic’s Claude Mythos and Claude Opus 4.8 and OpenAI’s GPT-5.5 to detect software vulnerabilities, raising concerns about what happens when that power becomes available.

Crypto investors were woken up to the increasing threat from powerful AI this week when Zcash developers revealed that Claude Opus 4.8 was enabled. getting a big risk which would have allowed the attacker to generate unlimited ZEC. As a result of network structurethere is no way to know if a fake ZEC has actually been created—and this uncertainty has driven ZEC to a premium. fall this weekend.

Experts warn that many more vulnerabilities could be discovered in the coming weeks and months as AI software becomes more sophisticated — and the devices become more accessible. Here’s a look at the growing threat, and how it’s already affected the crypto world.

Early forms of AI were used professionally as coding assistants, helping developers write, explain, and improve software. As technology advanced, researchers began to use the same systems for code analysis, software analysis, and risk analysis.

The shift from a coding assistant to a security tool coincided with a major shift in how AI is used within software development. After the implementation of the Claude Code in 2025, Anthropic report a significant increase in AI-generated code in its engineering teams, indicating a move from models that display code to systems that can write and run.

Security officials say the implications go beyond helping developers write code.

“AI is much better at reviewing code than most people and finding potential problems,” said Danny Jenkins, CEO and co-founder. ThreatLockerhe said Decrypt. Jenkins said that current AI systems are already accelerating vulnerability detection, while new models such as Mythos can expand this capability, calling it a “major challenge” at hand.

He said: “It’s only a matter of time before the bad guy is found.

According to Jenkins, AI is also lowering the barriers to entry in threat analysis, allowing more people to analyze code, identify vulnerabilities, and take action. As the availability of energy-efficient systems continues to grow, it is expected that risk-taking trends will increase.

“Pre-AI, cybersecurity threats are increasing every year,” he said. “Post-AI, it’s been very fast, and I think it’s grown fast for two reasons. One is that you can use AI to find vulnerabilities and exploits, and the number of people who have the ability to do that has grown exponentially. You don’t have to be a script kid now.”

As AI systems became more capable, companies began to use them for cybersecurity. Tuesday, Anthropic to be expanded access to Project Glasswing, giving 150 companies and organizations access to Claude Mythos to help identify and address issues with the software before it’s mass release.

In April, Mozilla later to be revealed that Anthropic models helped identify hundreds of bugs they fixed in the Firefox browser, while Calif. researchers used Mythos Preview in a project that produced one of the first public features. direct Apple M5 chips.

Stanislav Fort, a former researcher at Google DeepMind and Anthropic and now the founder and chief scientist of the security company. In passinghe said concerns about AI-driven risk exposure are legitimate, but often misunderstood.

“The counterintuitive response is to try to control the gates of power. I think this is a hidden security, and security and darkness is one of the worst ideas,” Fort said. Decrypt. “The possibility of a zero-day discovery has already been spread across the globe that no one can stop. Attempting to hide borders does not eliminate danger; it only slows down and diminishes the defenders who need these tools the most.”

Fort said the biggest risk is that defenders, especially open source developers, may lack access to advanced AI tools available to attackers.

“That imbalance is the real danger,” he said. “The answer is not prohibition; it is to democratize the self-defense group.”

Anthropic isn’t the only one pushing cybersecurity-related AI models. In May, Microsoft launched MDASHa threat detection system that the company said helped identify previously known Windows vulnerabilities.

The risk of crypto

Crypto and DeFi are starting to feel the effects of AI-driven virus hunting. Blockchain projects have always been attractive targets because there is so much money at stake and so much of the code is publicly available. Jenkins said that as AI begins to find software bugs, open-source crypto projects could become easy targets for security researchers looking for bugs and threat actors looking to exploit them.

One of the clearest examples of how advanced AI models can help researchers uncover vulnerabilities that have existed for years under human scrutiny is independent security researcher Taylor Hornby. to be revealed The biggest risk in Zcash’s Orchard’s secret pool that he got with the help of Claude. Opus 4.8.

The flaw would have allowed an attacker to create unlimited fraud ZEC, and it went unnoticed for years before it was built. It is not known if the weapon was used at this time.

“This vulnerability existed from Orchard’s launch in May 2022 until an emergency fix was deployed on June 1, 2026,” Shielded Labs, a Zcash development organization, wrote in a disclosure. “Due to the privacy of Orchard and the nature of the virus, there is no sure way to determine, using cryptography alone, whether such use occurred.”

The attack comes as DeFi protocols are already facing one of their worst years yet they take advantage of it. More than $840 million has been stolen from DeFi projects in the first five months of 2026, including more than $600 million in April alone in terms of service complexity. KelpDAOand Drift program.

The rise of the so-called ‘rip off vibe,’ where attackers use AI code to detect, steal information, create malware, and other tasks, has raised concerns that AI is lowering the barriers to cyberattacks.

According to Natalie Newson, chief blockchain researcher at the security platform Web3 CertiK, while April was particularly difficult for cryptos, the overall trend remains stable and less than the number of events seen in previous years.

“April 2026 was a bad month for using crypto currency; there were only three days without use when at least $10,000 was taken,” he said. “However, when we look at the bigger picture, the number of incidents (excluding fraud) is likely to have been consistent and still lower than the 2023 rate.”

Although AI makes DeFi easier to implement, according to Blockaid CTO Raz Niv, the biggest threat is not AI replacing hackers but augmenting them, allowing attackers to focus on more sophisticated methods while AI does routine work.

“The good news is that defenders can use the same tools,” he said. “AI-assisted analysis and simulation is critical for security teams trying to navigate.”