Bybit Says AI Security Blocked $700M in Threats in H1 2026, a Year After $1.46B Hack

Roughly 18 months after suffering one of the largest exchange breaches in crypto history, Bybit is claiming its rebuilt security infrastructure, anchored by AI-driven detection systems, prevented $700 million in losses during the first half of 2026.

The exchange made the announcement this week, framing the figure as evidence that its post-hack security overhaul is working. The $1.46 billion breach in early 2025 gutted Bybit's reputation and triggered industry-wide scrutiny of hot wallet management and multisig security practices. Now the platform is leaning on AI as its primary proof point that things have changed.

There is no independent verification of the $700 million figure. Bybit has not disclosed the methodology behind the calculation, the specific threat categories it covers, or how the exchange defines a "prevented loss" versus a blocked transaction that may never have succeeded regardless. That ambiguity matters. Quantifying averted damage is notoriously difficult even for traditional financial institutions with mature security programs. For a crypto exchange making this claim 18 months after a catastrophic breach, skepticism is the appropriate starting posture.

That said, the broader shift toward AI in exchange security is real and accelerating. Machine learning models have proven genuinely effective at flagging anomalous withdrawal patterns, identifying compromised API keys, and catching address-poisoning attacks before funds move. Several large exchanges have quietly deployed real-time behavioral analytics that can freeze suspicious transactions in milliseconds, faster than any human security team could respond. The question for Bybit is not whether AI security tools work in principle; it is whether the $700 million figure reflects rigorous internal accounting or is a marketing construct designed to restore user confidence.

"Bybit's AI-driven security advancements highlight the critical role of technology in mitigating financial risks and enhancing trust in crypto markets."

The framing of that statement tells you something. "Enhancing trust" is doing a lot of work in that sentence. Trust is rebuilt through transparency: published security audits, third-party penetration testing results, on-chain proof-of-reserves updates, and detailed post-mortems. A single unverified dollar figure, however large, does not substitute for any of that.

Context worth keeping in mind: the 2025 Bybit hack surpassed the $625 million Ronin Network exploit from 2022 and the $320 million Wormhole bridge attack, making it the single largest theft in crypto history at the time. The reputational hole Bybit is climbing out of is genuinely deep. That context also explains why the exchange has strong incentive to publicize any security wins, real or overstated. Exchanges that have suffered major breaches historically face prolonged user attrition; Mt. Gox never recovered, and even Bitfinex took years to stabilize volumes after its 2016 hack.

If Bybit's AI systems are performing as described, that is a meaningful development for the industry. Centralized exchanges remain the highest-value targets in crypto, and any credible improvement in detection capability reduces systemic risk. But "credible" is the operative word. Until Bybit publishes its threat detection methodology or invites an independent audit of its security claims, the $700 million number sits somewhere between genuine progress and a well-timed press release. The market should treat it accordingly.