Bybit Says AI‑Driven System Recovered $300 Million for Thousands of Users in 2025
The Singapore‑based exchange reported that its new risk‑management framework intercepted more than $300 million in fraudulent withdrawals and blocked three million credential‑stuffing attempts during the year, highlighting the growing role of artificial intelligence in crypto security.
Bybit’s 2025 Security Initiative
In a December 2024 announcement on X, Bybit disclosed the results of its 2025 Security Initiative, a program that pairs proprietary AI models with third‑party on‑chain analytics from firms such as TRM Labs, Elliptic and Chainalysis. According to the exchange, the system:
- Flagged $500 million in withdrawal requests for further review in Q4 2025.
- Successfully intercepted and returned $300 million to more than 4,000 users whose funds were targeted by impersonation and investment‑fraud schemes.
- Identified 350 high‑risk “pig‑butchering” addresses on‑chain, protecting roughly 8,000 accounts from potential loss.
- Automatically labeled 350 suspicious wallets and manually added another 600 via internal ticketing, averting an estimated $1 million of imminent fraud.
- Blocked over 3 million credential‑stuffing attempts linked to account‑takeover campaigns throughout the year.
David Zong, Head of Group Risk Control at Bybit, said the aim for 2025 was to turn “risk control into an active and intelligent guardian” by merging AI‑driven pattern detection with real‑time intelligence feeds. The firm’s statement emphasized that the technology not only shields its own users but also contributes to a broader mapping of fraudulent networks across the ecosystem.
Context: Crypto Fraud Remains a Major Threat
Chainalysis estimates that $17 billion in digital assets were lost to scams, hacks and other illicit activity in 2025. Bybit’s figures represent a modest slice of that total, but they illustrate how targeted interventions can dramatically reduce loss for individual investors. The exchange’s three‑tier risk framework—ranging from big‑data analytics for low‑risk anomalies to immediate withdrawal freezes for confirmed scam wallets—appears designed to balance security with uninterrupted trading for the majority of customers.
Industry Analysis
-
AI as a Front‑Line Defender – Bybit’s results reinforce a growing consensus that machine‑learning models, especially those trained on on‑chain transaction data, can spot emerging fraud patterns faster than manual review teams. The reported 3 million credential‑stuffing blocks suggest that automated credential‑checking tools are scalable and effective against large‑scale account‑takeover campaigns.
-
Collaboration with Analytics Vendors – Integrating feeds from TRM Labs, Elliptic and Chainalysis gives Bybit real‑time visibility into known malicious addresses. This “intelligence hub” approach may become a de‑facto standard for exchanges seeking to stay ahead of sophisticated threat actors who constantly adapt their tactics.
-
Quantifiable User Impact – Recovering $300 million for roughly 4,000 users translates to an average restitution of $75,000 per affected account, a figure that dwarfs the typical loss reported in many retail crypto scams. While the numbers are encouraging, the broader industry still faces a $17 billion shortfall, indicating that similar solutions need broader adoption to make a dent at the macro level.
-
Regulatory Implications – As regulators worldwide tighten AML/CFT obligations, exchanges that can demonstrate proactive fraud mitigation using AI may find it easier to satisfy supervisory expectations. Bybit’s public disclosure of its methodology could serve as a benchmark for future compliance frameworks.
- Potential Risks – Heavy reliance on AI carries the danger of false positives that could interrupt legitimate withdrawals. Bybit’s tiered alert system, which includes a one‑hour cooling‑off for high‑risk wallets, appears designed to mitigate this, but further data on user experience will be needed to assess the trade‑off between security and convenience.
Key Takeaways
- $300 M reclaimed: Bybit’s AI‑driven system returned $300 million to over 4,000 users, preventing large‑scale impersonation and investment‑fraud losses.
- 3 M credential‑stuffing blocks: Automated detection halted more than three million attempted account takeovers in 2025.
- AI + on‑chain intel: Partnerships with TRM Labs, Elliptic and Chainalysis power a real‑time risk engine that flags suspicious addresses and transaction patterns.
- Three‑tier framework: The model distinguishes low‑, medium‑ and high‑risk activities, applying proportional controls while preserving normal trading flow.
- Industry relevance: The initiative showcases a viable template for exchanges seeking to curb the $17 billion fraud scourge that plagued the crypto sector in 2025.
Bybit’s results suggest that sophisticated, AI‑enabled fraud prevention can deliver tangible financial protection for users. Whether other platforms will replicate this approach remains to be seen, but the exchange’s public reporting may accelerate broader adoption of similar technologies across the cryptocurrency ecosystem.
Source: https://cryptopotato.com/bybit-retrieves-300m-for-thousands-of-users-through-ai-enhanced-fraud-prevention-report/
