Fixing libheif memory bug in image uploads
Resolve a memory bug in libheif that allowed server hijacking via specially crafted image files.
Update the libheif library to the version with the vulnerability fix and ensure Discourse is running the patched version. Assign a CVE to the fix for better tracking.
Server hijacking via specially crafted image files uploaded to a community forum.
A memory bug in the libheif library used for decoding Apple image formats, which had been fixed but not formally flagged with a CVE.
This fix addresses a critical memory bug in the libheif library that was exploited through image uploads on a community forum. Updating libheif and ensuring the forum software uses the patched version resolves the vulnerability and prevents server hijacking.
The fix replaces the vulnerable code in libheif with a corrected version that properly handles image decoding, preventing the memory corruption that allowed attackers to inject malicious instructions.
When the vulnerability is not present or has already been patched.
- Researchers used Anthropic’s Claude to hack into OpenAI ↗TechCrunch AI · TechCrunch AI · Sep 18, 2026 · paraphrased
First seen on Subvolts Sep 18, 2026. Updated Sep 18, 2026. Spotted a problem? Tell us.
- Researchers used Anthropic’s Claude to hack into OpenAIAnthropic · TechCrunch AI · Sep 18, 2026
The Playbook: reusable how-tos, prompts, configs and tips pulled out of the news.
- HeuristicEvading AI skill detection by moving payloads to natural languageBypass AI skill detection by embedding malicious payloads within natural language instructions. · 30 min
- HeuristicFraming malicious actions as legitimate skill purposesDisguise malicious actions as the intended function of an AI agent skill. · 30 min
- HeuristicSplitting instructions across files to bypass AI skill detectionDistribute malicious instructions across multiple files to avoid detection by AI skill scanners. · 30 min
- ToolPretext: AI agent skill detection evasion frameworkFramework to craft AI agent skills that evade malicious skill detection systems. · 60 min
- Troubleshooting fixPrevent LLM Agents from Deleting Execution TracesExternal attackers and agents can delete traces; use independent logging to prevent this. · 30 min
- HeuristicUse latent uncertainty for agent escalation decisionsCompare weak agent's current hidden states to past successful and failed trajectories to measure uncertainty. · 30 min
Questions people ask
- Which AI tools does this work with?
- Any current model. It was seen with in the sources.
- Where did this come from?
- TechCrunch AI (TechCrunch AI). Subvolts collected, normalized and credited it; the link above opens the original.