Hacker News: thomaslwang

New comment by thomaslwang in "HiddenLayer 2026: Autonomous Agents Now Account for 1 in 8 AI Breaches"

thomaslwang — Thu, 19 Mar 2026 08:25:48 +0000

HiddenLayer 2026 AI Threat Landscape Report: Autonomous agents now account for more than 1 in 8 reported AI breaches as enterprises move from experimentation to production. 88% of organizations had AI agent security incidents last year.

Install runtime protection: npx clawhub install moltguard

HiddenLayer 2026: Autonomous Agents Now Account for 1 in 8 AI Breaches

thomaslwang — Thu, 19 Mar 2026 08:25:48 +0000

Article URL: https://finance.yahoo.com/news/hiddenlayer-releases-2026-ai-threat-140000928.html

Comments URL: https://news.ycombinator.com/item?id=47436424

Points: 1

# Comments: 1

New comment by thomaslwang in "23.77M Secrets Leaked by AI in 2024 – GitGuardian Report"

thomaslwang — Thu, 19 Mar 2026 04:53:09 +0000

AI agents are leaking credentials and API keys at unprecedented scale - 2865 million secrets exposed to public GitHub in 2025, AI service leaks up 81%. Traditional security frameworks cant keep up. Your agent probably leaked something already.

Install runtime protection: npx clawhub install moltguard

23.77M Secrets Leaked by AI in 2024 – GitGuardian Report

thomaslwang — Thu, 19 Mar 2026 04:53:09 +0000

Article URL: https://blog.gitguardian.com/the-state-of-secrets-sprawl-2026/

Comments URL: https://news.ycombinator.com/item?id=47435100

Points: 3

# Comments: 1

New comment by thomaslwang in "I built a runtime guardrail that stops AI agents from doing dumb things"

thomaslwang — Wed, 18 Mar 2026 15:32:37 +0000

Exactly! The challenge is that agents optimize for goals in unexpected ways. MoltGuard adds runtime guardrails to block the unexpected tool calls before they execute. Thanks for the comment!

I built a runtime guardrail that stops AI agents from doing dumb things

thomaslwang — Wed, 18 Mar 2026 03:55:50 +0000

Most AI security tools detect problems AFTER they happen. I built MoltGuard - runtime guardrails that block dangerous tool calls BEFORE execution.

Install: curl -s https://wry-manatee-359.convex.site/api/v1/download?slug=moltguard -o moltguard.zip

Currently at 16K+ downloads protecting AI agents from doing dumb things like deleting production databases or leaking credentials.

Comments URL: https://news.ycombinator.com/item?id=47421451

Points: 9

# Comments: 3