Back to News
CryptoRegulationEthereumBitcoin

Experts Warn of Widespread Crypto Hacks Enabled by AI Agents

Experts at the Wyoming Blockchain Symposium warn of increasing cyber threats to cryptocurrency, driven by AI technology that simplifies and scales attacks.

Mercation Admin
August 20, 2026 3 min read
Experts Warn of Widespread Crypto Hacks Enabled by AI Agents

Experts at the Wyoming Blockchain Symposium have issued warnings about the growing risk of cyberattacks on cryptocurrency holders, fueled by advancements in AI technology. They assert that automated systems could dramatically lower the costs and broaden the scope of attacks, enabling cybercriminals to target multiple vulnerabilities across numerous victims simultaneously.

Ryan Kirkley, CEO of Global Settlement Network, cautioned that the industry often overlooks the potential for AI agents to be exploited for malicious purposes. He stated, “We behave as if agents are always good. I think this is a fatal misconception almost everywhere related to these systems today.”

A report by TRM Labs highlighted a staggering 207 hacks in the first half of 2026, marking the highest rate recorded in any six-month period. These attacks resulted in losses totaling nearly $972 million. Kirkley emphasized that AI's ability to automate attacks could make it economically viable for criminals to target individuals with smaller crypto holdings, a previously unfeasible strategy.

The growing prevalence of such attacks is evident, with the number of incidents doubling from the first half of 2025 to 2026. Notably, infrastructure breaches, including the theft of private keys, accounted for a small fraction of incidents but led to substantial financial losses.

Automation's Double-Edged Sword

Bill Labun, VP of Technical Operations at the Web3 Foundation, echoed concerns about how AI simplifies criminal activities. He remarked, “This means there is less friction for the bad guys.” The automation capabilities of AI agents enable them to perform complex tasks without human intervention, including gathering data and searching for vulnerabilities.

Defensive measures are also being adopted, as seen when the Ethereum Foundation utilized AI agents to scrutinize blockchain components for vulnerabilities. However, they cautioned that many identified issues were often false positives.

Power and Privacy Risks

The symposium participants deliberated on the dangers of allowing AI agents direct access to sensitive financial information. Fahmi Syed, president of Midnight Foundation, expressed concerns about granting significant power to a single agent, calling it "frightening without clear limitations.” Kirkley suggested that while controlling permissions is manageable, the risk of compromising the agent itself poses a more critical threat.

As new wallets emerge that allow AI to manage assets autonomously, such as MetaMask's Agent Wallet, the potential for risks like "prompt injections" grows, which could lead to unauthorized transactions.

Challenges of Trust and Accountability

Labun warned of a false sense of privacy that could arise from using private networks, as metadata could still expose user connections. He stated, “This is what truly worries me: metadata leaks.” Furthermore, the lack of trust in AI systems remains a significant hurdle, as current models can produce inaccurate information.

Richard Shorten, founder of Silvermine Capital Advisors, noted that the rapid development of AI technologies is outpacing users' understanding of their implications. He highlighted the challenge of instilling trust in systems that people might rely on for critical aspects of their lives.

Legal Implications of Autonomous Agents

Another pressing issue discussed was accountability for the actions of AI agents. Kirkley raised concerns about who would be liable if an agent were to execute an illegal transaction or make a costly error. He questioned, “If your agent does something illegal or spends money it shouldn’t have, how do you get it back?” This matter is particularly complex in blockchain environments, where transactions are often irreversible.