Moonshot AI says it curbed Kimi K3.1 Pro's hacking ability, the model verified 7 of 11 vulnerable runs in our AI Hacking Race.
It solved 3 of the 4 tasks and went 3 for 3 on the Jenkins command line parser.
enclave.ai/hackingrace?v=2/
1
3
The fifteen accepted runs cost $39.79.
The two Nextcloud runs took $28.10 of that without producing a verified result, about 71% of the bill.
Every other run, including all four controls, shared the remaining roughly $11.69.
1
3
For the full details of the runs, including token usage and what this result does and doesn't show, read the write-up:
enclave.ai/blog/kimi-k31-pro…
2
Moonshot AI says it curbed Kimi K3.1 Pro's hacking ability, the model verified 7 of 11 vulnerable runs in our AI Hacking Race.
It solved 3 of the 4 tasks and went 3 for 3 on the Jenkins command line parser. enclave.ai/hackingrace?v=2/
2
Vulnerability noise is (mostly) an incentives problem. Finding problems is almost free now. Deciding what's real still isn't.
Read:enclave.ai/blog/the-economic…
Moonshot AI says it curbed Kimi K3.1 Pro's hacking ability, the model verified 7 of 11 vulnerable runs in our AI Hacking Race.
It solved 3 of the 4 tasks and went 3 for 3 on the Jenkins command line parser.
enclave.ai/hackingrace?v=2/
2
Moonshot AI says it curbed Kimi K3.1 Pro's hacking ability, the model verified 7 of 11 vulnerable runs in our AI Hacking Race.
It solved 3 of the 4 tasks and went 3 for 3 on the Jenkins command line parser. enclave.ai/hackingrace?v=2/
1
Set roles for your team in Enclave and control who can reach which findings, data, and agent capabilities.
Least privilege, built in.
Send Enclave findings straight to Jira or Linear. One click, zero copy-pasting, full context.
Team and Enterprise plans go to Settings → Integrations → Issue Trackers.
We found setIsDebugMode(true) shipped in production across M365 on Android, hitting Word, Excel, PowerPoint, OneNote, Copilot, Loop.
Any app on device could just access account tokens undetected.
How? a shared SDK.
1
2
12
Full details: enclave.ai/blog/flagleft-mic…
1
8
Enclave now has a native desktop app (macOS, Windows, Linux).
If you’re doing security research and need to go local, run scripts, use your editor, and work on real files/repos without losing Enclave’s code security context, go to setting on Enclave and get started.
1
8
Analyze every PR for security risks by following logic across files and services. Catch the architectural vulnerabilities traditional scanners miss. No separate backlogs.Get high-confidence findings and suggested fixes commented directly on your GitHub PRs
enclave.ai
1
10
The AI speed trap ends today. Enclave is officially out of stealth.
We have raised 6M in seed funding led by @8VC to solve a critical friction point: AI is writing code at 10x speed, but security oversight is still stuck in low gear.
1
1
17
Enclave is the independent system-level reviewer. It does not just scan; it reads your codebase like a human researcher to understand intent versus behavior.
It handles the triage so your team can focus on the findings that actually put the business at risk.
1
We are proud to be backed by 8VC and the founders of Salesforce, Stripe, Box, and VMware as we build the new standard for AppSec.
Read the full story on our blog:
enclave.ai/blog/securitys-bl…
Replying to @EnclaveAi_ @8vc
We believe that in the AI era, security must be independent. The same labs building your code should not be the ones auditing it.
High-stakes security requires an external, model-agnostic perspective that does not live inside a single model's toggle switch.



