# ZeroDrift makes AI message compliance models generally available

_Published Wednesday, September 23, 2026 at 10:21 AM EDT · AI · Latest · Tier 2 — Notable_

![ZeroDrift makes AI message compliance models generally available — Primary](https://images.siliconangle.com/blogs.dir/1/files/2025/03/Screenshot-from-2025-03-10-07-54-49.png)

ZeroDrift has made its Anchor 3.0 models generally available through an API that checks messages generated by AI agents before they are sent. The system can flag, rewrite or block a message, send it for human review and record the decision for an audit.

The three versions cover high-volume messages, regulatory checks and longer documents or company policies. ZeroDrift says its flagship model detected more than 90% of violations in a test based on financial industry rules. ZeroDrift published that benchmark itself, and the company has described the software as a way to reduce risk rather than catch every violation.

## Sources

- [siliconangle.com](https://siliconangle.com/2026/09/23/zerodrift-launches-three-models-for-real-time-ai-compliance-checks/)

---
Canonical: https://techandbusiness.org/newswire/4howm9_HqXkGsPtKvgMK4w
Published: 2026-09-23T14:21:21.708Z
Story chronology: 2026-09-23T13:00:48.000Z
Retrieved: 2026-09-23T15:52:34.369Z
Publisher: Tech & Business (techandbusiness.org)