Binance co-founder and co-CEO Yi He said on Nov. 4 that the exchange is taking outside criticism over its security and access-management practices head-on, laying out how it plans to handle user funds, explain the cause of the attack, and tighten defenses after the recent incident.
In a post on X at about 20:06 Taipei time, Yi He wrote that Binance is not a perfect company, but one that is still evolving. She said the criticism reflects the standards and expectations people hold for the exchange.
The post came after the Oct. 29 hot wallet incident and Binance’s preliminary investigation report released on Nov. 3. That report said about $6.34 billion in digital assets were affected, around $2.85 billion had been frozen or intercepted, roughly 98% of the affected tokens had been replenished, and Binance would absorb the financial gap. Rather than revisit the attack sequence, Yi He focused on how Binance is responding to criticism and what it plans to change next.
Yi He breaks the response into three questions
Yi He said Binance, as a trading platform, first needs to answer three issues clearly.
The first is whether user funds will be affected. She said Binance will fully take responsibility and framed that decision as part of the company’s contractual commitment. She added that, in legal terms, a sufficiently large funding gap could theoretically lead down a bankruptcy path, but Binance chose to handle the matter with users at the center.
The second is how the problem happened. Yi He said some people advised her to disclose fewer details before the preliminary report was published, but she chose to make the information public anyway. Her reason was simple: trust cannot be built without transparency. She said that if the report helps peers understand the complexity of social-engineering attacks and where guardrails should be drawn around AI permission boundaries, then publishing it serves a purpose.
The third is how to prevent a repeat. She said Binance has already started work on department-level permission management, security environments, device updates, and resetting AI tools. The company is also rebuilding its security boundaries under zero-trust and minimization principles, with a security audit report to follow.
At the end of the post, Yi He invited top AI security experts and technical talent to join Binance, saying candidates can message her directly or apply through openings listed on Binance’s website.
Security criticism centers on internal AI tools and DevOps access
Discussion around the incident has gone beyond the fact that Binance was attacked. Security firm Vectra AI said in its analysis that internal AI tools can act like hallways between teams within an attack chain. Once an attacker has passed identity checks, activity through those tools can resemble normal collaboration, without the attacker needing to exploit the tools themselves.
Vectra AI said lateral movement can cross multiple functional boundaries and end up scattered across different logs. In that setup, individual controls may each appear to be working while still failing to reveal the full attack chain.
The firm grouped the detection gaps into three points:
- Normal appearance: a 0-day is replaced by the use of legitimate tools;
- Successful authentication: signing follows a normal path;
- Invisible movement: lateral movement is hard to see end to end.
In plain terms, many controls may pass on their own, while no one is watching the layer where a verified user turns a tool into a bridge.
Similar questions have been raised around DevOps access. A commentary by TechBooky argued that if engineers’ laptops hold long-lived high-privilege tokens for tasks such as container deployment, the production path is effectively tied to a single endpoint. It said access should shift to least privilege and just-in-time, or JIT, access. Machine identities and service tokens should not remain on developer laptops for long periods, and high-value operations should also be subject to separation of duties, allowlists, and quota controls.
Community discussion has landed in much the same place: endpoint protection did not stop the 0-day used in the interview assignment, while deployment permissions sat too close to the wallet-signing path.
Binance ties zero-trust changes to a new hiring push
For users, Yi He’s post drew a hard line on who will bear the financial burden and listed zero trust, least privilege, device updates, and AI tool resets as measures already in motion. That moves the post-report trust repair effort from apology toward process and controls.
For the industry, her decision to speak directly about AI permission boundaries turns the "hallway" problem flagged by Vectra AI and DevOps circles into a public checklist of issues the exchange says it will address.
The talent angle is direct as well. Yi He publicly called for AI security specialists and said that joining Binance and testing the security system firsthand would be the best way to validate it.
The next points under scrutiny are when the security audit report will be released, and whether the boundaries around department permissions and internal AI tools can actually shift from collaboration-first convenience to visible trust boundaries.


