WatchTower builds governance and runtime security tooling for AI systems, and we also use AI in producing some of our own content. This page explains what that means for you.
1. AI-assisted content
Some articles, summaries, and illustrative graphics on this Site are produced with AI assistance and reviewed by our team before publication. AI systems can produce inaccurate or outdated statements. Verify anything you plan to rely on for security, legal, or compliance decisions.
2. Illustrative product output
Dashboards, risk scores, detections, alerts, and telemetry shown on the Site are simulated demonstrations, not live monitoring of real customer environments.
3. Probabilistic systems and human oversight
AI detection and policy-enforcement features are probabilistic. They can produce false positives and false negatives. WatchTower is designed to support — not replace — human review. Customers should keep a human accountable for consequential decisions, consistent with the NIST AI Risk Management Framework and the EU AI Act's human-oversight expectations.
4. No professional advice
AI-assisted output on this Site is not legal, medical, financial, or regulatory advice.
5. Data used with AI
Do not submit sensitive personal data, regulated health data, payment card data, secrets, or classified information into public forms or demos. We do not use information submitted through our waitlist or investor forms to train third-party foundation models.
6. Responsible AI commitments
- Document intended use and known limitations of AI features.
- Keep audit-grade logs so AI-driven actions can be reviewed after the fact.
- Provide policy controls and kill-switch style enforcement for autonomous behavior.
- Test for prompt injection, data exfiltration, and unsafe tool use before release.
7. Contact
Questions about our AI practices: info@watchtoweragents.com