More than a decade after the infamous Mt. Gox collapse, its former CEO Mark Karpelès is revisiting history — this time with the help of artificial intelligence. By feeding the exchange’s original 2011 codebase into Anthropic’s Claude AI, Karpelès has uncovered what modern tools might have identified as the fatal weaknesses behind one of crypto’s earliest and most catastrophic failures.

In a detailed analysis shared on X (formerly Twitter), Karpelès revealed that Claude labeled the code as “feature-rich but critically insecure,” pinpointing the precise vulnerabilities that led to Mt. Gox’s first major breach — just months before the exchange’s complete collapse in 2014.

“I didn’t get to look at the code before taking over; it was dumped on me as soon as the contract was signed,” Karpelès admitted. “I know better now — due diligence goes a long way.”

Source: Mark Karpelès

Karpelès acquired Mt. Gox from its original developer, Jed McCaleb, in March 2011, only to face a devastating hack within three months that drained around 2,000 Bitcoin — worth roughly $230,000 at the time but valued at over $200 million today.

Claude AI’s forensic breakdown of Mt. Gox’s flaws

According to the AI-generated report, Mt. Gox’s early system was a paradox: technically impressive but dangerously fragile. Claude’s findings highlighted:

  • Weak administrative security — including retained access for prior admins even after the ownership transfer.
  • Poor password hygiene — with one critical admin account using a weak password.
  • Minimal internal documentation, making oversight and patching nearly impossible.
  • Lack of proper network segmentation, allowing a breach in one part (like a WordPress blog) to compromise the entire exchange.

The AI identified the root cause as a “sophisticated but avoidable” chain reaction that began when attackers compromised Karpelès’ WordPress and social media accounts. That breach gave them the access needed to exploit Mt. Gox’s internal vulnerabilities.

Some security patches implemented before the hack — such as switching to salted password hashing, fixing SQL injection flaws, and adding withdrawal limits — helped prevent an even greater loss. Without these improvements, Claude estimated that tens of thousands of BTC could have been stolen.

“The salted hashing prevented mass compromise and forced individual brute forcing,” the AI noted, “but no algorithm can protect weak passwords.”

Still, while AI could have flagged many of these issues in real time, the report emphasized that the human factor — poor operational security and lack of oversight — played an even greater role in the downfall.

Mt. Gox’s ghost still haunts the crypto market

More than 10 years later, Mt. Gox’s impact continues to ripple through crypto markets. The exchange’s court-appointed trustees have been gradually repaying creditors, with around 34,689 BTC still awaiting distribution ahead of the October 31, 2025 repayment deadline.

Despite concerns of mass sell-offs, previous repayment rounds have not triggered major volatility. Analysts suggest that most creditors are long-term holders or have already hedged exposure.

Karpelès, who has long maintained his innocence, said the AI analysis serves as both a historical lesson and a technological thought experiment:

“If tools like Claude or ChatGPT had existed back then, maybe we could have avoided one of crypto’s first major disasters,” he reflected.