Open-source AI risk tools rising
A new roundup showed open-source AI platforms now analyze both A-share and US markets for high-dimensional portfolio risk assessment, enabling rapid anomaly detection and simulated macro shocks. These tools are shortening prototyping cycles — but the brief flags the need for human validation and cross‑testing against traditional risk metrics.
FinRobot’s GitHub repo shows 6.4k stars and a codebase of ~289 commits for an AI agent platform that combines LLMs, reinforcement learning and quant toolkits for automated equity research and risk tasks [GitHub]. (github.com) A peer‑reviewed FinRobot paper details a Chain‑of‑Thought equity‑research design and agent experiments for valuation and anomaly detection published on arXiv in Nov 2024 [arXiv]. (arxiv.org) The Open Source Risk Engine (ORE) maintains production‑grade XVA, PFE and market‑risk sensitivity modules and published a new Python package release on PyPI on Feb 5, 2026 to expose ORE bindings for downstream risk pipelines [PyPI]. (pypi.org) Post‑trade firm writeups and ORE docs cite adoption by large derivatives shops and describe ORE’s QuantLib base and CRIF‑formatted sensitivities used by “1,000+” major derivative trading institutions for benchmarking and validation [Wilmott/Acadia]. (wilmott.com) OpenBB published an integration case on Nov 6, 2025 showing an AKShare/Tushare extension that adds China A‑share and Hong Kong endpoints to OpenBB’s AI workspace, while FinRobot’s issue tracker notes it currently lacks built‑in A‑share market feeds (issue raised 2025) [OpenBB blog][FinRobot issue]. (openbb.co) The Risk Atlas Nexus package (PyPI v1.0.4, released Nov 14, 2025) and the AI Risk Atlas paper describe a taxonomy and tooling that use ontologies/knowledge graphs to link model vulnerabilities and mitigation artifacts for open governance of AI risk [PyPI][arXiv]. (pypi.org) Robust Intelligence’s open AI Risk Database was turned into an open‑source dataset in collaboration with MITRE to help track supply‑chain and adversarial risks for public models, with the announcement and GitHub rollout reported in 2025 [Robust Intelligence / MITRE]. (executivebiz.com) OpenBB’s extension framework plus ORE’s Feb 2026 Python package and FinRobot’s agent tutorials form concrete roadmaps for prototyping: OpenBB’s AKShare/Tushare adapters (Nov 2025) and ORE’s Python bindings (Feb 5, 2026) are explicit integration points for feeding A‑share/US market data into backtests and scenario engines. (openbb.co)
