SHOW HN:Employees use ChatGPT.Your CISO has no visibility. EU AI Act says fix it

yohann_senthex — Thu, 07 May 2026 16:26:37 +0000

Article URL: https://senthex.com/en/

Comments URL: https://news.ycombinator.com/item?id=48051331

Points: 1

# Comments: 0

Show HN: Senthex AI firewall for LLM API calls (one line of code, 16ms overhead)

yohann_senthex — Tue, 07 Apr 2026 08:15:27 +0000

Article URL: https://senthex.com:443/proxy/

Comments URL: https://news.ycombinator.com/item?id=47672143

Points: 2

# Comments: 0

New comment by yohann_senthex in "Show HN: Dochia – automated API testing for agentic build-test-fix loops"

yohann_senthex — Thu, 02 Apr 2026 12:53:44 +0000

Solid approach on the agent readable JSON. The real value here isn't just "catch bugs faster" it's that agents can now run comprehensive negative testing as part of the build loop, which is increasingly critical with agentic workflows.

One angle worth exploring: injection attacks. When Dochia fuzzes your endpoints, does it specifically test for prompt injection vectors if the API backend is an LLM? Most security tools miss that because they're generic but if an agent is writing an API that wraps Claude or GPT, you'd want fuzzing that includes things like "what if the user input gets templated into a system prompt?"

Same with token limits if the fuzzer generates massive payloads, does it distinguish between "API can't handle large inputs" vs "you'll bankrupt the user with token costs if this endpoint is exposed"?

Might be niche, but seems like low hanging fruit in the agentic testing space.

Hacker News: yohann_senthex

SHOW HN:Employees use ChatGPT.Your CISO has no visibility. EU AI Act says fix it

Show HN: Senthex AI firewall for LLM API calls (one line of code, 16ms overhead)

New comment by yohann_senthex in "Show HN: Dochia – automated API testing for agentic build-test-fix loops"