Cardano founder Charles Hoskinson has responded to reports that attackers withdrew roughly 4,000 Bitcoin, worth approximately $320 million, from the Liquid Network federation wallet. The reported loss prompted Liquid to pause network activity while federation members reviewed the incident.

Hoskinson linked the breach to broader concerns about artificial intelligence and software security. His response centered on whether traditional testing methods can keep pace as AI tools become increasingly capable of finding code weaknesses and automating attacks. The incident is one of the larger losses involving Bitcoin-linked infrastructure in recent years, and cross-chain and bridging services have repeatedly been targeted across the industry, making the security of the underlying software stack a recurring point of debate.

How the Incident Unfolded

Liquid said the attackers moved the funds through SideSwap’s Peg-out Authorization Key (PAK) service. However, both Liquid and SideSwap stated that the PAK itself was not compromised. SideSwap instead pointed to a reported vulnerability in Elements, the Bitcoin-based software that underpins the Liquid Network. Liquid is a Bitcoin sidechain operated by a federation of member organizations, and Elements is the open-source codebase on which that network is built, which is why a weakness in the software itself, rather than a compromised key holder, became the central focus of the post-mortem.

We are aware of a security incident on @Liquid_BTC . Purported white-hat hackers have withdrawn 4,000 BTC ($320 million) from the Liquid Federation wallet. The @Blockstream team is working on contacting them on-chain with a signed message. What we know so far is that the funds… — Liquid Network (@Liquid_BTC) September 6, 2026

Following the transaction, Liquid disabled bridge nodes and paused network operations. Exchanges received alerts and began suspending, or preparing to suspend, L-BTC deposits and withdrawals while the investigation continued.

The transaction also carried an OP_RETURN message from the attackers, who described themselves as white hats and asked to be contacted through the blockchain. Liquid has not indicated that this claim changes its investigation, and the network continues to treat the transfer as an unauthorized movement of federation-held Bitcoin, which remains under review. How the situation is resolved, and whether the Elements report is confirmed, will likely shape how other federated and sidechain networks review their own dependency on shared open-source code.

Formal Methods Enter the Security Debate

Hoskinson argued that formal methods could offer stronger protection as AI-driven security risks grow. Formal methods use mathematical specifications and proofs to verify whether software follows its intended design before developers release them. This approach contrasts with conventional testing and audits, which examine selected scenarios and code paths rather than proving properties across a system’s behavior.

AI systems can now review source code, search for weaknesses, and automate parts of phishing and social-engineering attacks. The reported Coldcard theft of approximately $130 million in Bitcoin has also drawn attention to the role advanced tools may play in future crypto attacks.

Cardano’s Focus on Formal Verification

Cardano has incorporated academic research, peer review, and formal verification into its development process. Its Ouroboros consensus protocol originated from formal research, and Cardano developers have used mathematical methods to test important protocol properties.

The network also uses Haskell and Plutus, which rely on strong type systems and functional programming. These tools can help developers identify certain software errors earlier. Hoskinson has long presented this development model as a way to reduce weaknesses before code safely reaches production. Whether the wider industry adopts similar rigor in response to AI-era threats remains an open question, and the outcome of the Liquid investigation is likely to feed that discussion.