Liquid Network Suffers $320 Million Bitcoin Breach Amid AI Suspicions
Roughly 4,000 bitcoins worth $320 million were stolen from the Liquid network, sparking speculation over whether advanced artificial intelligence like OpenAI's GPT-6 Astra was used.
An estimated 4,000 bitcoins, valued at approximately $320 million, were withdrawn from a wallet belonging to the Liquid network in a security breach this week. According to reports, the funds were transferred using a mechanism designed to move Bitcoin out of the network, prompting Liquid to suspend new transactions. Beyond the question of who is behind the hack, another intriguing issue has emerged: Was advanced artificial intelligence involved in identifying the vulnerability?
The Timing and the Launch of GPT-6 Astra
The suspicion stems largely from the timing. Just three days before the funds were withdrawn, OpenAI launched GPT-6 Astra, a new model described by the company as its most advanced regarding cyber capabilities. According to OpenAI, Astra is its first model to reach a 'Critical' level in cyber proficiency. This means that, when combined with appropriate tools and permissions, the model can identify previously unknown security vulnerabilities and develop ways to exploit them in protected systems without requiring human guidance at every step.
OpenAI published data illustrating the extent of this capability's improvement. In internal testing, Astra successfully identified two zero-day vulnerabilities and exploited them. The company stated that it is working to disclose these vulnerabilities to the parties responsible for the affected software. In other tests, the model managed to find weaknesses in hardened systems and construct attack vectors that could potentially allow system takeover.
The Caveat: No Direct Proof of AI Involvement
However, a crucial caveat remains: there is currently no evidence that Astra was involved in the Liquid breach. Protos, which published one of the primary investigations into the incident, noted that various sources raised the possibility that the hacker utilized AI, partly due to the close proximity to Astra's launch. Nevertheless, the site could not determine whether the vulnerability exploited by the attacker was indeed discovered by an artificial intelligence model, nor is there any public indication that the hacker specifically used Astra.
This incident illustrates the profound shift transforming the cybersecurity landscape. In the past, a hacker attempting to uncover a system vulnerability required deep technical knowledge, time, and occasionally an entire team. Today, advanced models are capable of performing portions of this work independently, ranging from code analysis to searching for points of vulnerability. While this capability can assist researchers in finding and fixing flaws before they are exploited, it can also be leveraged by attackers. OpenAI itself warns that the rise in Astra's cyber capabilities necessitates stronger defensive measures.
At this stage, therefore, there is no proof that the chatbot assisted in extracting hundreds of millions of dollars from Liquid. Yet, the mere fact that this possibility is deemed plausible enough to spark a debate highlights the monumental shift taking place beneath the surface: the war between hackers and defense systems may rapidly evolve into a conflict between artificial intelligence models.
