SpotBlock has been watching Ethereum mainnet 24/7 and labeling what it sees. spotblock.ai is where we explore what that data makes possible — for security research, for models, and for the people who need fresh threat signal.
Every number on this page is traceable to the spotblock threat database.
Accumulated by the SpotBlock scanners since June 2026 — public on-chain data, labeled and reviewed by humans.
Directions we are currently evaluating. This is a map of what the data supports — not a product list.
Fresh, labeled feeds of scam families, deployer clusters and funding-taint chains for wallets, exchanges and security tooling. Freshness is the point — not static dumps.
Human-verified verdicts and structured audit findings — the scarce ingredient for training and evaluating smart-contract security models.
Contract-risk classification built on the labeled corpus — small, cheap models that classify at corpus scale and complement frontier-model analysis.
Reproducible per-vulnerability-class benchmarks, so "detected" means a measured rate — not a score without evidence.
Drafting detectors, invariants and fuzz harnesses with LLMs, validated by static analysis and fuzzing — a loop that finds new vulnerability shapes, not just known ones.
Funding-taint tracing, actor clustering and mimic detection via bytecode/source embeddings — catching mass-deployers and campaign rotation.
Exploration phase — these are possibilities we are evaluating, not commitments, and nothing here is for sale yet. Label quality comes first: measured precision against human verdicts gates everything above.
If you work on smart-contract security, threat modeling or ML for on-chain data, we'd like to hear what you'd want from this data. Reach us through the SpotBlock contact page or watch the live platform.
SpotBlock