Security Research Program
Building Safer AI, Together with CygnusAlpha
We operate AI systems in production—specifically GenAI—which is genuinely uncharted territory. We’re working That’s why we need you. We believe transparency and collaboration with the security community is the only
Our Philosophy
with non-deterministic systems where frameworks are immature and vulnerabilities are still being discovered.
We don’t pretend to have all the answers.
responsible path forward. This program is our commitment to that principle.
Program Scope
In Scope
Out of Scope
Safe Harbor
We commit to:
We ask that you:
AI Security Pioneer Award: Annual recognition for researcher who contributed most to
What We Offer
Recognition Tier System
🥉 Bronze (Low)
🥈 Silver (Medium)
🥇 Gold (High/Critical)
🏆 Platinum (Novel)
Special Recognition:
advancing GenAI security understanding (not just finding bugs, but advancing the field).
Engagement & Response Framework
Submission Process
Where to Report:
engage@cygnusalpha.com
[SECURITY] Brief description
[your-site]/security-pgp
What to Include:
Our Response Timeline
| Timeframe | Action |
|---|---|
| 24 hours | Initial acknowledgment + case ID |
| 72 hours | Preliminary assessment + severity classification |
| 7 days | Detailed response with remediation plan |
| 45 days | Target resolution |
Handling Process
Step 1: Initial Contact (Within 24h)
Step 2: Assessment (Within 72h)
- Reproduce the vulnerability
- Classify severity using AI-Security adapted CVSS:
- Critical: Data exfiltration, complete jailbreak, credential access
- High: Partial data leakage, authorization bypass, harmful content generation
- Medium: Information disclosure, context manipulation
- Low: Minor prompt injection with limited impact
Step 3: Detailed Response (Within 7 days)
Step 4: Resolution & Recognition
- Fix deployed
- Offer recognition options
- Request feedback on our handling
- Ask permission for public disclosure
Dispute Resolution
If a researcher disagrees with our severity assessment:
- We will schedule a 30-min video call to discuss.
- We will bring in a neutral third party from our advisory board to mediate.
- We will document the rationale transparently and upgrade the tier if warranted.
Q: Why participate if bounties are small? Q: What if I find something critical but really need cash? Q: How do you determine severity in AI systems where impact is fuzzy? Q: What happens if we disagree on disclosure timeline?
FAQ
A: GenAI security is where cybersecurity was in the 1990s—wide open. We’re offering co-authorship,
recognition, and learning opportunities in the hottest security domain.
A: Talk to us. We’ll be honest about constraints but will work with you—payment plans, consulting exchanges,
or connecting you with clients who might hire you.
A: Great question. We use adapted CVSS but add AI-specific factors. We’re transparent about our reasoning and
open to discussion. This is new territory for everyone.
A: We commit to 45 days standard, extensible by mutual agreement. If we can’t agree, we’ll defer to a neutral
security researcher from our advisory board.

