ZeroDrift launches Anchor 3.0 for real-time AI compliance

ZeroDrift is not a newcomer to this problem. In an August preview, it disclosed that Anchor combined deterministic checks with an open-source model trained on regulatory material and attorney-labeled communications, and at that time it described the software as a risk-reduction layer rather than a guarantee that every violation would be caught. The production release now splits that idea into three sizes: Mini for high-volume traffic, the flagship for applying more than 200 prebuilt rules across FINRA, the Securities and Exchange Commission and other regulations, and Max for long documents, attachments and a company's own policies without extra fine-tuning. All three were post-trained from Google LLC's Gemma E4B and Alibaba Group Holding Ltd.'s Qwen3.8-27B, so the family leans on open base models rather than starting from scratch.

The models sit inside ZeroDrift's wider compliance platform, which intercepts communications and can flag, rewrite, block or route problematic messages for human review while recording the decision for later audit. That routing choice matters: the sales pitch is not that AI replaces reviewers, but that it decides which messages ever reach one. CEO Kumesh Aroomoogan framed the difficulty as running capable agents inside a regulated business, where the check has to happen every time before anything goes out.

The open question is how much weight those numbers can carry. The benchmark's attorney-labeled data was produced independently by Surge AI Inc., but ZeroDrift published the benchmark itself, so the performance figures remain company claims. Whether regulated firms treat Anchor as a first-pass filter or as the last line before a customer sees a message is likely to hinge on how those claims hold up under outside testing.