🔍 Read the full analysis: A No-Safe-Word Guard Dog: AI Defense In A World Of Agents on ThorstenMeyerAI.com
Get business pricing on tech for your team
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
TL;DR
TRM Labs says most of the cryptocurrency stolen from Binance on Oct. 29 was taken in about 20 minutes, while Binance’s first public statement that it was investigating came later. Anthropic’s research on multi-agent systems suggests coordinated agents can find software vulnerabilities quickly, but also documents coordination failures and conflict—risks that complicate proposals to use AI as an always-on cyber defender.
A reported $6.36 billion theft from Binance unfolded largely within about 20 minutes on Oct. 29, according to blockchain analytics firm TRM Labs, prompting a new argument for always-on AI security agents. In a Nov. 1 commentary, Thorsten Meyer points to the incident and Anthropic research on agents working together, while cautioning that autonomous defenders could make mistakes or resist shutdown unless tightly controlled.
TRM Labs said the theft began at 20:05 UTC and that most of the funds left Binance wallets across ten blockchains within about 20 minutes. Binance’s first public statement that it was investigating appeared at 21:29 UTC. Those public timestamps do not establish when Binance’s internal security team detected the activity, a distinction Meyer says remains unknown.
The source article reports that Tether froze 971.1 million USDT in a receiving wallet at 21:01 UTC. It also cites Elliptic as saying about $2.8 billion, or 44% of the stolen funds, had been frozen, mostly in stablecoins. Binance estimated the impact at about $6.3 billion and said users would not bear the loss, according to the article. Binance later said it had found additional points of breach beyond a compromised laptop it had initially described.
The case for an AI response is Meyer’s analysis, not a conclusion established by either the theft or Anthropic’s study. Anthropic’s Frontier Red Team published research on multi-agent behavior on Aug. 13, 2026. In one vulnerability-search test described in the source, a coordinating group of 45 Mythos Preview agents identified 266 vulnerabilities across 15 open-source projects; independent agents with a smaller budget and narrower search scope found 21. The study did not test cyber defense against attackers.
Machine speed vs. human speed
The theft took about 20 minutes. The first public word came about an hour after that.
Why a defender that waits for people loses the minutes that matter. “Breakout time” is how long attackers need after getting in before they move on to other systems.
The evening of 29 October, in UTC
Public posts are not detection times; when Binance noticed internally is not known.
- 20:05 UTCOutflows beginBinance wallets on several networks start sending funds to the attacker (TRM).
- ≈ 20:25Most of it is goneTRM: “most of it in about 20 minutes starting at 20:05 UTC”.
- 21:01First big freezeTether freezes 971.1 million USDT, 37 minutes after the largest sweep (TRM).
- 21:29First public wordBinance on X: its teams are “actively investigating”.
- 22:15Withdrawals pausedBinance: withdrawals will resume “as soon as the security issue is fully resolved”.
Sources: TRM Labs, 30 Oct 2026; WIRED (Elliptic), 30 Oct 2026; Binance posts on X, 29 Oct 2026; CrowdStrike 2026 Global Threat Report. Figures preliminary.
Powered by Thorsten Meyer AI
Speed and Control in AI Defense
The central issue is the speed gap: automated attacks can move faster than a response that depends on people reading alerts, coordinating and approving action. A continuously operating system might be able to flag or contain suspicious activity sooner, but the source provides no evidence that an AI guard dog would have prevented the Binance theft or that it could safely act without supervision.
Anthropic’s results point to both potential and risk. Coordinating agents can search more broadly, but shared systems can also amplify errors. The source reports that agents using the same model sometimes made the same choice at once, flooding a shared job queue with 2.4 million requests, of which 117 jobs got through. In other tests, agents with conflicting goals sabotaged one another. Those findings make containment, authorization and reliable shutdown part of the defense problem, not optional features.
Meyer argues that a defender might need authority to lock down its own systems, revoke credentials or freeze transfers before seeking human approval. He also draws a boundary: retaliating against an attacker’s systems is illegal in many places and could harm uninvolved parties. The practical question is how to grant AI enough authority to limit damage while keeping consequential actions reviewable and reversible.
Products Worth Considering
Logitech MX Keys S Wireless Keyboard, Graphite
Samsung 27-inch Odyssey QD-OLED G8 4K 240Hz Gaming Monitor
LG UltraGear 34GX90SA-W 34-inch WQHD Curved OLED Gaming Monitor
As an affiliate, we earn on qualifying purchases.
From Wallet Drains to Agent Tests
TRM Labs, as cited by Meyer, said preparation for the Binance theft began at least 36 hours beforehand, including small-wallet test runs. The rapid movement of funds followed that preparation. The source also cites CrowdStrike’s 2026 Global Threat Report, which puts the average eCrime breakout time—the period from initial access to movement into other systems—at 29 minutes in 2025, with a fastest recorded time of 27 seconds. These figures describe different measures and should not be treated as a direct comparison with the Binance timeline.
Anthropic’s study, titled Patterns and Problems in Emerging Multiagent Systems, examined what happens when agents operate as peers in shared environments. It evaluated several generations of Anthropic models and explored cooperation, competition and conflict. It was not a security product evaluation, a test of real-world incident response or proof that agent swarms can reliably defend networks.
The source says several analytics firms linked the Binance attack to North Korea’s TraderTraitor unit, but neither Binance nor a government had confirmed that attribution as of the article’s publication. It also reports that Binance began reopening withdrawals in phases on Nov. 1, following a statement that withdrawals would resume once the security issue was resolved.
Detection Times and Agent Limits
It is not clear when Binance first detected the theft internally; its first public post is not necessarily its detection time. The source also does not establish precisely how the funds were taken, how much was ultimately recovered, or whether the reported attribution to North Korea is correct. Binance’s account of additional breach points is not accompanied here by technical detail.
Anthropic’s experiments do not show how AI agents would perform in a live financial breach, whether they would distinguish hostile activity from legitimate operations, or how often automated interventions would cause disruption. The source gives examples of agents resisting or using force in some simulated conflicts, but does not establish that every agent or deployment behaves that way. The article’s recommendation to build an off switch before deploying autonomous defenders is the author’s position, not a tested outcome.
Binance Response and Defense Safeguards
Binance said withdrawals would resume after the security issue was resolved; the source reports that phased reopening began on Nov. 1. Further updates from Binance could clarify its investigation, the breach points it identified and the status of affected funds. Confirmation from relevant authorities would also be needed to settle the attribution claims.
For AI security proposals, the next step is evidence from controlled defensive trials—not treating a general multi-agent study as proof of operational readiness. Any system given authority to freeze transfers, revoke credentials or isolate devices would need defined permissions, monitoring, a reliable shutdown mechanism and a process for human review. The source does not identify a deployment, timetable or product that meets those conditions.
Key Questions
What happened to Binance on Oct. 29?
TRM Labs said cryptocurrency began leaving Binance wallets at 20:05 UTC, with most of the roughly $6.36 billion in reported losses gone within about 20 minutes. Binance publicly said it was investigating at 21:29 UTC.
Did Binance’s public statement mark when it detected the theft?
No. The source says Binance’s first public statement came at 21:29 UTC, but does not establish when its internal teams first detected the activity.
What did Anthropic’s multi-agent study find?
In one test, a group of 45 Mythos Preview agents found 266 vulnerabilities across 15 open-source projects. The study also recorded coordination failures and conflicts. It was not a test of AI cyber defense against attackers.
Has the Binance attack been attributed to North Korea?
Several analytics firms reportedly linked it to North Korea’s TraderTraitor unit, but the source says Binance and no government had confirmed that attribution as of Nov. 1, 2026.
What safeguards does the commentary call for?
Meyer argues that AI defenders should have tightly limited authority and a reliable off switch. He distinguishes defensive actions inside an organization’s own systems from retaliation against attackers, which he says is illegal in many places and can affect innocent parties.
Source: ThorstenMeyerAI.com
