NEAR AI suffered a digital breach when an unknown attacker gained unauthorized access to its internal systems. The company immediately locked down its platform, completely pausing its NEAR AI Cloud inference service to stop the bleeding. The artificial intelligence firm quickly assured the public that the core NEAR Protocol blockchain remained untouched, meaning no cryptocurrency was stolen.
We recently identified unauthorized access to NEAR AI internal systems. As a precaution, we have paused NEAR AI Cloud inference while we investigate.
Inference on NEAR AI Cloud runs inside hardware-isolated secure enclaves (TEEs) to keep customer prompts private.
NEAR Protocol and other NEAR services are not affected.
We will share an update within two hours.
— NEAR AI (@near_ai) November 9, 2026
They also stated that user prompts were safely locked inside hardware-isolated secure enclaves, protecting private customer queries from the attackers.
NEAR AI confirmed they had successfully closed the backdoor, fixed the broken servers, and replaced the compromised passwords. The cloud service is now back online.
Update: On November 8, an attacker used a vulnerability in SGLang, the inference engine we run, to access part of our infrastructure. We’ve closed the access and fixed it.
Affected users have been contacted. If you haven’t heard from us, no action is needed.
https://t.co/iuRUo7dBW4— NEAR AI (@near_ai) November 10, 2026
However, social media users immediately attacked the company’s choice of words. On X, @LazyICP pointed out a massive loophole, questioning what would happen to all the other customer data if only the prompts were kept private.
@near_ai Customer prompts private…. what about all the other data?
— Lazy (@LazyICP) November 10, 2026
Meanwhile, Jeffrey on X mocked the company for calling the breach “unauthorized access” instead of admitting they were hacked, urging users to switch to an ICP subnet.
@near_ai Aaaah the new name for HACKED is unauthorized access 🤣🤣🤣 . Stop with that nonsense and switch to an ICP subnet
— Hold The Line (@jeffrey17249) November 10, 2026
The danger of centralized servers in crypto
Many digital asset companies build a blockchain to handle their money but still use traditional cloud providers to run their daily software. This creates a severe structural weakness. If a hacker finds a weak password or a flaw in the server code, they can walk right in.
Investors assume that because a company created a highly secure cryptocurrency token, every other product they build shares that same unbreakable security. However, the blockchain only protects the money itself. The moment a company builds an artificial intelligence tool or a customer dashboard, they often revert to renting cheap, fast servers from traditional tech giants to save money.
Critics online point directly at networks like ICP because they offer a completely different structure. On X, Marco Zucaro, noted that AI hackers are becoming a daily threat because many crypto companies still rely on traditional Web2 cloud servers instead of true decentralized networks like the Internet Computer Protocol (ICP).
@near_ai Practically it got hacked!!!
Dom Williams ICP announced this 2 years ago… AI hackers were going to become a daily thing… that’s because NEAR like the rest are web 2 fragmented ledgers…the Blockchain tech is applied to record coin movements, that’s all. Go STUDY ICP— Marco Zucaro (@zucaro_mar83047) November 10, 2026
An Internet Computer Protocol subnet does not rely on a traditional cloud server. Instead, it spreads the actual software and data across thousands of independent computers worldwide. There is no central server for a hacker to break into. The entire application runs directly on the blockchain itself. When crypto companies rely on old Web2 servers to host their new AI tools, they expose modern financial technology to outdated security risks.
A blockchain might record token movements, but if the website attached to it lives on a weak centralized server, the overall ecosystem remains highly vulnerable to everyday cyberattacks.
Why secure AI prompts do not protect customer data
The assurance that prompts stayed private leaves a blind spot regarding the rest of the system’s data. When you sign up for a cloud service, you provide your email address, billing information, digital wallet connections, and usage history.
A hardware enclave only protects the specific mathematical task happening inside it at that exact moment. It does not protect the massive databases storing all the other customer details. When hackers breach internal systems, they usually target these exact databases. Even if the criminal cannot read the specific questions you asked the AI yesterday, they might still steal your personal identity, contact information, or login credentials.
Once a hacker steals a list of customer emails, they cross-reference those addresses with known cryptocurrency exchange accounts. They craft highly realistic fake emails pretending to be the hacked company, offering a fake security patch or a fake refund. When the terrified user clicks the link, the hacker drains their actual crypto wallet. Hiding behind the safety of an enclave prompt ignores the reality that losing basic contact data often leads to massive financial losses for regular retail investors down the road.
In another development, hackers stole approximately $133 million from Crédit Mutuel Arkéa, a major French banking group. The attackers did not break through the bank’s main security walls. Instead, they hacked into a smaller third-party service provider that worked for the bank.
Enjoyed this? Bookmark DeFi Planet, explore related topics, and follow us on Twitter, LinkedIn, Facebook, Instagram, Threads, and CoinMarketCap Community for seamless access to high-quality industry insights.
Take control of your crypto portfolio with DEFI PLANET PRO, DeFi Planet’s suite of analytics tools.
