MIT's HardFlow Makes AI Follow Strict Safety Rules

MIT researchers built HardFlow, an algorithm that forces existing AI models to respect non-negotiable rules, like safety limits and laws of physics, MIT News reported Sept. 14, 2026. It targets safety-critical jobs where output must stay high-quality and never break those requirements.