
NEAR AI Restores Cloud Service After SGLang Security Breach
NEAR AI has restored its Cloud service after a security breach that exposed account information belonging to around 100 users and data from 18 artificial intelligence agents, the company confirmed in a security report published on November 9. The incident began on November 8, when an attacker exploited a remote code execution vulnerability in SGLang, an open-source inference engine running on NEAR AI’s servers, to gain access to internal systems. The company said it moved quickly to fix the flaw, replace affected credentials and complete a forensic review, and that its blockchain protocol and other NEAR services were not affected. The breach is a reminder that the infrastructure behind the fast-growing market for autonomous AI agents is still maturing, and that a single flaw in a shared software component can put user data and agent activity at risk. It also arrived in a week when NEAR-based tokens were linked to the movement of funds stolen in the Binance exchange hack, a separate episode that NEAR AI has been careful to distinguish from its own incident.
What happened on November 8
According to the company’s security report, published on November 9, the attacker gained access to NEAR AI’s systems through SGLang, an open-source inference engine that the company used on its servers to run large language models. A remote code execution vulnerability is one of the most serious kinds of software flaw: it allows an attacker to run their own commands on a target machine from a distance, without needing a username or password. Once inside, the attacker was able to reach internal systems and access information that should have stayed private. NEAR AI said it fixed the vulnerability, replaced affected credentials and completed a forensic review of what had been touched.
What data was exposed
The breach exposed account information belonging to approximately 100 users, including email addresses and names. That is the kind of personal data that can be used for phishing attempts, where attackers send convincing messages pretending to be a trusted service in order to steal passwords or funds, so affected users will need to stay alert even though the direct exposure was limited. Eighteen legacy AI agents also had their chat histories, settings and logs accessed. For users who run agents that hold conversations, store preferences or keep records of past tasks, that is a meaningful intrusion, because an agent’s history can reveal what its owner asked it to do, what tools it connected to and what instructions it was given. NEAR AI said affected users were contacted directly, which is the right response when personal data is involved, since people cannot protect themselves against a risk they do not know exists. The company did not report any loss of customer funds, and there is no indication in the report that wallets or on-chain assets were touched.
How SGLang fits into NEAR AI’s stack
SGLang is an open-source inference engine, the software that sits between a large language model and the applications that use it, turning requests into generated text as efficiently as possible. That shared foundation is also a shared risk: a vulnerability in a widely used library can affect every service that depends on it, and attackers actively study popular open-source projects for weaknesses. NEAR AI’s experience shows how that risk plays out in practice. The flaw was not in NEAR AI’s own code or in the NEAR blockchain, but in a third-party component running on its servers. For the growing number of crypto projects that combine blockchain infrastructure with AI services, the lesson is that security depends on the whole stack, including the open-source tools that rarely appear in marketing material but do much of the heavy lifting behind the scenes.
The response: sessions revoked, tokens replaced, Private Chat offline
NEAR AI’s response included revoking active sessions and administrative credentials, replacing agent tokens and taking Private Chat offline. Each of those steps addresses a different way an attacker might keep access after the initial break-in. Revoking active sessions forces anyone who was logged in, including the attacker, to sign in again. Replacing administrative credentials closes off the privileged accounts that control the system. Replacing agent tokens cuts the connection between compromised agents and the services they talk to, so stolen tokens cannot be reused to impersonate an agent. Taking Private Chat offline was the most visible step for users, since it temporarily removed a feature people rely on, but it was also the most cautious: if a private messaging feature might have been exposed, shutting it down while the investigation runs is safer than leaving it open. Together, the measures describe a textbook incident response, focused on closing every door the attacker might have used rather than only the one they entered through.
NEAR Protocol and other NEAR services were not affected
NEAR AI was clear that the NEAR Protocol itself and other NEAR services were not affected by the breach. That distinction matters because NEAR AI is one part of a much larger ecosystem built around the NEAR blockchain, and confusion between the two could easily shake confidence in the network itself. The blockchain protocol, which secures transactions and runs smart contracts across a decentralised network of validators, is a separate system from the cloud servers that host NEAR AI’s agent services. An attacker who compromises a company’s servers does not automatically gain any ability to alter a blockchain, move tokens or rewrite transaction history. The company also said its blockchain protocol remained unaffected, with no established connection to the Binance hack. For token holders and developers building on NEAR, the practical message is that the network kept running normally throughout the incident, and that the damage was confined to the AI cloud service and the user and agent data it held.
A separate story: NEAR tokens and the Binance hack
The incident occurred during the same week investigators identified NEAR-based tokens being used to move funds associated with the Binance exchange hack. Blockchain analytics firm Elliptic reported that attackers created six temporary memecoins and used decentralised liquidity pools to transfer approximately $430,000 in stolen cryptocurrency. Researchers identified roughly 117,000 NEAR routed through pools involving tokens created between November 5 and November 7. The technique is a form of laundering: by creating new tokens, trading them against stolen funds in liquidity pools and moving the proceeds between connected wallets, criminals can make illicit money look like ordinary trading profit. It is important to keep the two events clearly separate. NEAR AI says there is no established connection between the breach of its Cloud service and the Binance hack, and nothing in the reported facts links the attacker who exploited SGLang to the group moving stolen Binance funds. The only real connection is timing and the shared use of the NEAR network, which is a public blockchain that anyone, including criminals, can use.
Why AI agent security matters more than ever
The breach lands at a moment when AI agents are moving from demonstrations to real products that hold conversations, manage tasks and connect to wallets and other services. An agent that remembers what its owner asked, which tools it used and what it decided is a rich target: its history can reveal personal information, business plans and financial intentions. The incident also shows how quickly the AI crypto sector is becoming interconnected. A flaw in one open-source inference engine can reach a cloud service, which can reach user accounts and agents, which can reach the wallets and applications those agents serve. As more projects build agent economies on top of blockchains, security reviews will need to cover the AI layer as seriously as the smart contract layer, because users will judge a platform by its weakest link, not its strongest.
What affected users should do now
NEAR AI said affected users were contacted, and anyone who received that message should treat it as the starting point for a few practical steps. Anyone whose email address was exposed should be extra cautious about messages claiming to come from NEAR AI or related services, and should never enter credentials or seed phrases through a link in an email. Users who ran one of the 18 affected legacy agents should review what those agents stored, including chat histories and connected settings, and consider whether any sensitive instructions or information need to be changed. The broader lesson applies to every AI crypto platform, not just NEAR AI: when agents hold personal data, the security of the underlying infrastructure deserves the same attention as the security of the blockchain itself.
Frequently asked questions
What caused the NEAR AI security breach?
An attacker exploited a remote code execution vulnerability in SGLang, an open-source inference engine used on NEAR AI’s servers, gaining access to internal systems on November 8. NEAR AI fixed the vulnerability, replaced affected credentials and completed a forensic review.
What data was exposed in the breach?
Account information belonging to approximately 100 users, including email addresses and names, was accessed. Eighteen legacy AI agents also had their chat histories, settings and logs accessed. Affected users were contacted by the company.
Was the NEAR blockchain affected?
No. NEAR AI said the NEAR Protocol and other NEAR services were not affected, and that its blockchain protocol remained unaffected. The breach was confined to NEAR AI’s Cloud service and the data it held.
Is the breach connected to the Binance hack?
NEAR AI says there is no established connection between the breach and the Binance hack. The two events happened in the same week, but nothing in the reported facts connects the two incidents.
What did NEAR AI do in response?
The company revoked active sessions and administrative credentials, replaced agent tokens, took Private Chat offline, fixed the SGLang vulnerability and completed a forensic review of the affected systems.
Source: crypto.news
—
—
—
Related reading on MM Crypto Analysis Web
- Best AI Crypto Coins November 2026: Five Projects Worth Watching
- Tokenised Money Market Funds Lead Finance Into the AI Agent Era
- AI Coding Agents Drive a 14x Surge in Bitcoin Payment Inquiries
More in AI Crypto on MM Crypto Analysis Web.
Pingback: AI Crypto Tokens November 2026: Sector Leads Market