Resect AI

AI Safety · Enterprise AI · AI Observability

Launch

Resect AI exits stealth with $25M to catch hallucinations mid-generation

Most rivals detect hallucinations by scanning finished text; this system intervenes inside model activations before a wrong answer is generated.

• Resect AI came out of stealth on September 3, 2026 with $25 million in funding from unnamed private-equity investors to build what it calls an accountability layer for enterprise AI.
• Founded by CEO Kevin Owens, Chief AI Officer Tim Walton, COO Tyler Gerber and CPMO Tommy Lofgren, the Washougal, Washington-based startup is building tools to observe, audit and modify how large language models generate answers.
• Instead of checking a model's finished output like most monitoring tools, Resect's patented tech injects probes into a model's internal activations during inference to catch fabrication signals before text is produced.
• The approach grew out of the founders training their own high-factuality model, then applying that reinforcement-training method to open models including DeepSeek, Qwen and Llama.
• The company employs about 30 people across the Seattle area, California, New York and Texas, with plans to grow to roughly 50 by the end of 2026; no valuation was disclosed alongside the $25M.
• Resect is entering a crowded AI observability market, but it's placing itself deeper in the model stack than typical wrapper-style monitoring tools that only review completed outputs.
• AI hallucinations reportedly cost businesses tens of billions of dollars a year, and enterprises in regulated fields like healthcare and finance won't move AI into production without a way to audit and correct model behavior in real time.