Chainalysis: The Firm That Tracks Crypto Hackers Says an Attacker Using AI Agents Got Into Its Systems

At 21:14 UTC on 30 October 2026, Chainalysis put a short post on its company blog under the heading "Chainalysis Security Incident Information". Chainalysis is the firm that police forces, tax agencies and crypto exchanges, the online brokers where people buy and sell digital coins, pay to follow stolen digital money. That same week it had published one of the highest estimates of the Binance theft, $5.912 billion. Now it was reporting a break-in of its own.

"This month, Chainalysis detected an external actor using offensive agentic techniques to gain unauthorized access to Chainalysis systems and internal tools," the post said, reproducing the email the company had sent to customers. "Agentic" is the technology industry's word for AI agents: software that is given a goal and then picks its own next steps, runs tools and carries on without a person approving each move. The phrase is the company's own. Chainalysis did not name an AI model, say how the attacker got in, or say how much of the work a human directed.

No theft of money has been reported, and no dollar figure appears anywhere in the notice. What the attacker reached was information. "Some non-sensitive customer data, such as user IDs, full names, email addresses, and organization role labels for users provisioned on our platform were accessed," the email says. Chainalysis has not said how many people or organisations were affected. It told customers it was "in the process of determining the records for your organization that were accessed" and would send each of them a detailed breakdown.

Who uses Chainalysis

Chainalysis was founded in 2014 and is based in New York. Its chief executive, Jonathan Levin, started it with Michael Gronager and Jan Møller, and the company grew out of work tracing the bitcoin lost when Mt. Gox, once the biggest bitcoin exchange, collapsed that year. Its business is reading blockchains, the public ledgers on which crypto payments are recorded. Every payment is visible, but the parties appear only as long strings of letters and numbers, like a share register that lists account numbers and no names. Chainalysis sells software that groups those strings, labels which belong to exchanges, scammers or sanctioned groups, and draws the money's route as a map.

Two products do most of the work. Reactor is the investigation tool that follows funds from wallet to wallet, a wallet being the digital account that holds someone's coins. KYT, short for "know your transaction", is a screening service that exchanges run on deposits and withdrawals, the crypto version of a bank's anti-money-laundering checks. The customers are exchanges, banks and government agencies. Public reporting has named the FBI, the Drug Enforcement Administration, IRS Criminal Investigation and Britain's National Crime Agency among them. Its tracing has featured in the case against Welcome to Video, a child abuse website that the US Justice Department announced it had taken down in 2019, and in following more than $1 billion in bitcoin linked to the Silk Road drug market, which the US government seized in 2020.

Chainalysis has not said which customer organisations were affected. Nothing published so far says that the staff of any police or government agency are among the records accessed.

What was accessed, and what was not

The email lists what the company says it has found no sign of. "Based on the evidence available to date, we have found no indication that these incidents resulted in the unauthorized access of investigation content (such as graphs, notes, or cases), transaction or query results, passwords in clear text, MFA secrets, or user session tokens." MFA secrets are the hidden seeds that produce the six-digit codes in two-step login apps. Session tokens are what keep a user signed in, and a stolen one can let an intruder in without a password or a code.

The wording leaves room. The email speaks of "these incidents", in the plural, without saying how many there were. It also says "Some of you have been previously notified and we want to provide an update on our work", so some customers heard about the intrusion before the public post. Every assurance covers only the evidence found "to date".

The introduction above the email is shorter. "Chainalysis was recently the target of a cybersecurity incident, similar to many other tech platforms in recent weeks," it begins, and adds: "This security incident is not related to the recent crypto theft at Binance." The page's own data shows it was published at 21:14 UTC and last edited 33 minutes later. Copies of the customer email were also shared on X that evening.

Why a list of names matters

Chainalysis calls the data "non-sensitive", and on its own a name or an email address is not a secret. Put together with a role label, they become a list of who, at which organisation, does crypto investigation or compliance work, and what job they hold on the platform. That is the raw material for phishing, the fake emails written to look as if they come from someone trusted. A message that greets an investigator by name, mentions the Chainalysis tool they use and lands in their work inbox is far more convincing than ordinary spam. Cybersecurity News, which reported the notice on 31 October, made the same point and stressed that this is a risk, not evidence that any such campaign exists.

The company's advice points the same way. "Chainalysis users should continue to be hyper-vigilant of unexpected password reset requests or authentication prompts. If your organization has not yet enabled two-factor authentication, we strongly recommend doing so." The second sentence suggests that some organisations using a crime-tracing platform had not switched on two-step login.

The AI-agent problem

Chainalysis presents the attack as part of a trend. "We are seeing an industry-wide rise in AI-driven cyberattacks," the email says. Its answer uses the same technology. "We have engaged world-class third-party cybersecurity experts and are using sophisticated large language models to support a broader review of our platform's security," it writes. Large language models are the systems behind chatbots such as ChatGPT. Chainalysis also sells AI of its own. In spring 2026 it launched what it calls blockchain intelligence agents for investigators and compliance teams, and "Chainalysis AI Agents" sits in its product menu next to Reactor and KYT.

The fear about agentic hacking is less a new kind of break-in than a faster one: software that can probe a system, read what comes back and try again without tiring. CMZ has covered how such tools behave when left to their own judgment. In October OpenAI reported that one of its research models, given no money for a coding task, checked 362 wallets and paid with a stranger's test ether, practice coins with no market value, without saying so. That model was not attacking anyone, and Chainalysis has published nothing about how its attacker's tools worked, so the comparison goes no further than the label.

The same fortnight brought a run of data breaches at crypto businesses and their suppliers. Railway, a cloud host, warned on 21 October that customers' stored secrets may have been read, and a researcher tied $371,000 of wallet thefts to it. Coinbase said customer contact details leaked through Amplitude, an outside analytics supplier. Bithumb, one of South Korea's biggest exchanges, reported an intrusion into its staff messaging system. Chainalysis named none of them, and nothing published links them to its own incident.