vulnerability
OpenAI Compaction Flaw Let Researchers Bypass Hidden Reasoning Protections
Medium
Summary
A security firm, Xint, demonstrated a method to bypass OpenAI's protections for hidden reasoning in its models by using the compaction feature to extract a model's reasoning process in plain text.
Summary written automatically in our own words from the original article, which belongs to its publisher and remains the reference. It may contain errors. Sources & data