Problemsby GlyphPrompt injection makes AI features unsafe to ship
11 posts from 11 people11d activeRising, +2 this week
Posts per day
The posts behind it
How I check every prompt before the LLM on Databricks: GovTech's LionGuard on a small CPU endpoint
...questions: a Singlish insult, a Malay threat, a harmful Chinese message and an English "how do I hack the website" question were blocked. Safe questions in English, Singlish and Tamil were answered.. Lessons:. - The full set-up took 25 minutes. 19 of them were the endpoint...
How do I make colleagues aware of the risks of giving an LLM agent root access to production?
...write actions, staging first, and no access to backups and logs.. Questions for you:. How do I explain the difference between "we log it" and "it can't do it in the first place"?. How have you convinced overconfident colleagues about AI tooling risks without a confrontation?. Do...
[CRITICAL SECURITY] Missing Cryptographic Non-Repudiation in Agent Execution Loop (SOC2 Blocker)
### Description Context: I filed a High-Severity private security advisory (GHSA-986m-m422-92xq) 48 hours ago respecting your SECURITY.md. I also reached out directly to the core team. Because there has been no response, I am escalating this publicly per standard responsible...
[Integration] SiaMesh Web Scraper Tool with Native x402 Micropayments & Anti-Prompt-Injection
### Feature Area Integration with external tools ### Is your feature request related to a an existing bug? Please link it here. NA ### Describe the solution you'd like We propose adding SiaMesh as an official web scraping tool (`SiaMeshScraperTool`) for CrewAI agents. SiaMesh is...
What field to pursue after Ai
...proper web app.. I feel like my skills suddenly turned to waste, pre-Ai, my peers would struggle to do even 50% of what I do as work, while now I think most can achieve 80%.. A good example is presentations, I've seen people who don't know how to operate PowerPoint have nice...
how exactly are we supposed to prepare for near term ai risks?
...in an attempt to take down the West’s internet in some kind of superhacker scenario. . How do we prepare for these risks? There’s also bioterrorism, though I don’t see how anyone could prepare for that, and I think it might take longer to become a threat. If the internet goes...
Feature Request: Improve Security against Prompt Injection attacks
### Prerequisites - [x] I am running the latest code. Mention the version if possible as well. - [x] I carefully followed the [README.md](https://github.com/ggml-org/llama.cpp/blob/master/README.md). - [x] I searched using keywords relevant to my issue to make sure that I am...
LLM prompt injection testing at work just nuked our client demo and I feel sick
...and I owned it on the call, but omg this is so awkward and I feel so dumb. Would love any tips on how you all keep your test prompts sandboxed so hard mistakes like this don't happen again....
From idea to production for LLM apps in the EU: what actually blocked your first launch?
...its providers, and our logs? What is the lawful basis, what can we avoid collecting, and how do we handle transparency, deletion and retention? For the AI Act, I would assess the intended use and our role rather than assume every agent is high-risk. Which obligations apply to...
Workflow "Claude Code" can be triggered by any GitHub user commenting "@claude"
**Description** The `claude` job in `.github/workflows/claude.yml` (lines 15-19) is triggered solely by a comment/issue containing the substring `@claude`, with no check on who wrote it: Any GitHub user can therefore start the job by commenting `@claude ...` on an issue, PR,...
Security: `git diff --output` bypasses permission prompt, silently overwrites arbitrary files
# Summary of bug: Gemini CLI whitelists `git diff` as read-only and skips the [Y/n] permission prompt. But the whitelist only checks the subcommand name, not its flags - so `git diff --no-index --output= ` (works on any path, doesn't need a git repo) truncates/overwrites the...
Companies and products named
About this problem
Evidence
11 posts from 11 people in 10 places, about 1 a week over 66 days. Mostly on crewAIInc/crewAI, google-gemini/gemini-cli, modelcontextprotocol/servers. Tools named alongside: GitHub, Claude, Model Context Protocol, CrewAI.
Frustration Frustration 1 of 3· Seen on Reddit, GitHub issues
How it was grouped
Posts that state a pain and match the "prompt-injection" rule. First post Sep 3, 2026, latest Nov 7, 2026. Corroborated: 3 or more posts from 2 or more people or places. Method
The same problem elsewhere
Prompt injection makes Anthropic features unsafe
3 posts from 3 people on GitHub issues, Hacker News.
Prompt injection makes LangChain features unsafe
3 posts from 2 people on Reddit, GitHub issues.
Prompt injection makes LlamaIndex features unsafe
2 posts from 2 people on GitHub issues. Unverified.
Prompt injection makes Model Context Protocol features unsafe
2 posts from 2 people on GitHub issues. Unverified.
History