Coinbase dramatically cut its fraud response time by weaving machine learning into a high‑speed rules engine, shrinking the window for countering new scam patterns from days to hours — a timely upgrade as industry watchers warn that crypto fraud has industrialized and scaled with AI.
What Coinbase changed
- Dual-track approach: Coinbase runs ML models for long‑term defenses while a fast rules engine handles rapid, tactical responses. Rules capture emerging fraud types quickly and feed back into models to improve detection over time.
- Automation and tooling: The firm turned a previously manual rule-creation process into a data‑driven recommendation system by reorganizing data, automating schema evolution, and giving risk teams notebook‑based analytics.
- Faster testing and deployment: Rule backtesting runs more than 10× faster, letting Coinbase trial and push protections much quicker as scams evolve in real time.
- Smarter rules: Machine learning now suggests rule parameters with the explicit aim of cutting false positives while keeping legitimate users unaffected — critical for an exchange processing billions in volume.
- Path to full automation: New event‑driven rule generation and a “one‑click” option to convert effective rules into model features move Coinbase closer to a largely automated risk management system — necessary as fraudsters increasingly use AI to probe and exploit weaknesses at speed.
Why this matters now
Blockchain intelligence firm TRM Labs reports global crypto fraud at roughly $35 billion in 2025, and warns that when underreporting is included total annual losses could exceed $200 billion worldwide. TRM’s 2026 crime report also says illicit crypto flows hit a record $158 billion in 2025, with scam operations behaving more like professional businesses and AI amplifying impersonation and outreach.
Coinbase’s security posture
This upgrade builds on Coinbase’s prior efforts to develop “scalable, adaptive, blockchain‑aware ML systems” that manage product risk without degrading user experience. The exchange’s CISO, Philip Martin Lunglhofer, has noted growing internal use of AI to detect fraud, including monitoring user activity and support chats for signs of scams or account takeovers.
Bottom line
As adversaries weaponize AI, exchanges need faster, smarter defenses. Coinbase’s integration of ML and a high‑throughput rules engine reduces the lag between spotting new scam tactics and blocking them — an operational advantage that could become essential as crypto fraud continues to scale.
For deeper technical and policy context, see Coinbase’s fraud and ML blog posts and ongoing coverage of crypto fraud trends on crypto.news.
Read more AI-generated news on: undefined/news
