Most AI products fail the same three security tests. The system prompt can be extracted in under ...Most AI products fail the same three security tests. The system prompt can be extracted in under ...
The network for creativity
Join 1.25M professional creatives like you
Connect with clients, get discovered, and run your business 100% commission-free
Creatives on Contra have earned over $150M and we are just getting started
Most AI products fail the same three security tests.
The system prompt can be extracted in under 10 minutes.
The guardrails block obvious jailbreaks but miss indirect prompt injection through documents, RAG sources and tool outputs.
There are no runtime evidence logs, so when a regulator or enterprise client asks "prove your model behaved", there is nothing to show.
I work on exactly this. That's my product AI Prompt Shield in the screenshot below, a guardrail API with runtime evidence logging mapped to OWASP's agentic security guidance and the EU AI Act. I have also published a book on LLM security and write about AI governance and compliance.
What I offer here on Contra, all fixed scope, no hourly meters:
LLM security and prompt injection audits
System prompt hardening and prompt engineering
AI governance documentation packs aligned to the EU AI Act
Technical writing on AI security and compliance
If you are shipping anything with an LLM inside it, get it tested before a user or a competitor does it for free. DM me.
#aisecurity #llm #promptengineering #aigovernance #euaiact #compliance #promptinjection #technicalwriting
Post image
Back to feed
The network for creativity
Join 1.25M professional creatives like you
Connect with clients, get discovered, and run your business 100% commission-free
Creatives on Contra have earned over $150M and we are just getting started