Jaswanth Alkur
Chittoor, India
Bio
Updated 10/10/26 · Provided by member · VerifiedIndependent AI safety researcher in India working on the monitors meant to catch a misbehaving AI agent: whether they can be fooled, whether they actually run, and whether their evidence is really independent. Since November 2025 I have written 21 papers with code and data anyone can rerun; one was accepted at the NeurIPS 2026 TAE workshop and about ten are under review. I audited 39 open source AI safety systems, and two maintainers shipped fixes after my reports. I pre-register my experiments and publish what fails.
Links
Updated 10/10/26 · Provided by member · Verified- Personal Website
- https://github.com/Alkur123
- LessWrong
- jaswanth-alkur
Projects
Agent monitors that cannot be talked down
Team Member
Grants
Updated 10/10/26 · By grantmaking.aiNo grants recorded.