Stopping a cyberattack while walking your dog - defensive AI security CEO says it's not ruff to do
What happened
Corma, a defensive-AI startup, says its agent models can detect and block live attacks and reported a $60 million seed round. The company ran vendor-run tests where offensive runs implanted persistent backdoors in a majority of scenarios while detection rates were much lower, which makes the vendor claims operationally real but derived from controlled exercises. Watch for independent validation or pilot results that confirm those response improvements in buyer environments
Why the category manager should care
Treat the vendor as a maturing category signal; require pilot-based proof points and clear rollback/authorization clauses before accepting live automated blocking
Key facts
- Company announced $60 million seed funding
- Vendor test matrix ran dozens of model-vs-model engagements and reported implant vs detection
- Early enterprise deployments cited by the vendor claim reduced response times