BLA releases hardening tool against AI reverse engineering agents

BLA has released a set of hardening tools for software authors and security teams, designed to prevent AI reverse engineering agents from extracting the core logic of binary, JS, and Python programs. Its methods include resource exhaustion, decoy payloads, and self-sealing mechanisms. BLA states that after hardening, each analyzer triggers deterministic resource limits and simulates 300×–2,800× token amplification; 14/14 built-in benchmark targets pass input-output equivalence tests. This tool is for defensive use only and must be installed from the official channel qybaihe/bla or the npm package bla-agents. Two features are disabled by default: in-band prompt decoys require adding --llm-directives to enable; signed attestation requires generating a manifest with --manifest, then signing it with --sign-manifest, and verifying the ed25519 signature and artifact SHA-256 via bla verify-manifest.

AI overview based on source reports · Updated 2 days ago