Security at Omega
Omega builds and tests software for companies that can't compromise on where their code lives. These are the controls behind that.
Isolation
On Omega cloud, each organisation gets a dedicated cluster. No two organisations share a cluster. When policy requires that code never leaves your network, we deploy Omega inside your own infrastructure.
Sandboxed evaluation
Models you evaluate run in isolated sandboxes, separate from each other and from the rest of the platform, so one run cannot affect another.
Red-teaming
Every product built on Omega is red-teamed before release. If your organisation has its own red team, they run the engagement on the Omega platform. If not, Omega's security team does it, only on the products you put in scope.
Access
- Access is by named seat. Each user has their own account, and seats cannot be shared.
- Your organisation decides who gets a seat and can remove access at any time.
- Omega staff access customer content only when needed to provide the service, such as an Omega-led red-team engagement you have authorised, or a support request you raise.
Data protection
- Data is encrypted in transit between your browser and Omega.
- Your code, models and specifications are never used to train models.
- Customer content is deleted within 30 days after a subscription ends, or sooner on request.
Compliance
Omega is in pre-release and has not yet completed a third-party security audit. Enterprise customers can request answers to their security questionnaires, and we will share our plans for formal certification as they are set.
Report a vulnerability
If you find a security issue in Omega, email Ravi Shankar at ravishankar@svyasachi.com with "Security" in the subject line. Please give us reasonable time to fix the issue before disclosing it publicly. We will not take legal action against good-faith research that follows these guidelines.

