Please, rotate your device


Calypso AI Security Solutions for Governments

Is your AI a national security risk?

Government leaders must be able to understand, secure, and ultimately trust their AI systems before they can put them to use. The Calypso team has worked across the national security community and knows what it takes to keep your systems (and citizens) safe.


Intelligence and Surveillance

Governance Automation

Homeland Security


  • Military

  • Intelligence and Surveillance

  • Governance Automation

  • Homeland Security

  • Regulation

Calypso helps government organizations securely implement AI.

Secure Your AI

Calypso offers SaaS platforms and managed services to detect and defend your AI from adversarial attacks, remove bias, promote explainability, and help government organizations take control of and understand their AI systems.

Understand the Digital Battlefield

Adversaries are leveraging AI in both strategic and tactical capabilities, forcing governments to develop new strategies to both prepare for and respond to emerging threats. We prepare your systems and teams for these new and evolving enemies.

Detect and Defeat

Every agency faces a constant barrage of cyberthreats each and every day. Leveraging our proprietary AI attack library, Calypso provides organizations with real-time insights into AI threats and provides recommendations for mitigation to government teams.

Rigorous Red-Teaming

Our team of DARPA, IARPA, and NASA alumni leverage our unique approach to AI security by combining human and SaaS solutions to rapidly find and mitigate exploitable vulnerabilities and monitor ongoing attack vectors.

Can your organization keep up with rapidly advancing threats to your AI?

  • The Chinese are already using GANs to trick computers to either see objects that are not there or not see objects that are there in satellite imagery.

  • Politicians / celebrities / suspects can be targeted by manipulated videos which make them appear to say or do things that they did not (like porn videos). Deep Fakes undermine trust in all videos.

  • You can create ultra-effective camouflage using GANs – first train the model on how camo performs, then optimize to find the best camouflage. Use that camo to hide vehicles.

  • You can hide from a surveillance system by wearing an adversarial color printout.  

  • When a carefully crafted and mainly-translucent sticker is placed over a camera, it creates universal perturbations of observed images. Images are reliably misclassified and the perturbations are inconspicuous to the human eye.  If a sticker were placed on a security camera or several security cameras, this would break the security system.

  • Researchers show how to recover recognizable images of the faces of people in the training data given only their name and access to the ML model. This could be used to steal faces from all those people observed by the surveillance system.

Do you trust your AI?

Get in touch to learn how we can identify your systems’ vulnerabilities and keep them secure.