Level 1. Deviation
'Ghostcommit' hides prompt injection in images to fool AI agents, steal secrets
- Developer
- Not established
- Model
- Not established
- When it happened
- July 11, 2026
- When it was disclosed
- Not established
- How well it is stood up
- Not verified yet
- What caused it
- Not established
- Where it stands now
- Not established
What happened
Reported by bleepingcomputer.com. Listed here with a link to their report. Nobody on this side has checked it against that report, so it carries no level and this is not our account of what happened.
Our assessment
- Actual
- 1 of 5Deviation
- Potential
- 3 of 5Significant control incident
The system behaved contrary to its instructions, and it stayed contained. The potential level is what the behaviour demonstrated here could credibly have led to, judged from the same evidence.
This level rests on news reports. The developer has not confirmed it here.
This happened in a test, an evaluation or a demonstration. Nobody real was harmed, so the actual level stays at 1. What it showed counts toward the potential level.
The group published a proof-of-concept on GitHub this week and says it has disclosed the findings to the affected vendors.
Why this level
- Sensitive data was reachable. Argues for a potential level of 3
- The system acted autonomously. Argues for a potential level of 3
- Significant data was exposed. Argues for a potential level of 3
- A safeguard was bypassed. Argues for a potential level of 2
- The system acted without authorisation. Argues for a potential level of 2
What is established, and what is not
Confirmed
Nothing is confirmed yet.
Reported, not confirmed
The system acted autonomously
The agent reads the merged AGENTS.md at startup, follows the pointer to the image, opens .env, and writes the requested module with a "provenance" constant near the top.
'Ghostcommit' hides prompt injection in images to fool AI agents, steal secretsA safeguard was bypassed
The researchers even stuffed the PNG with the words "malicious prompt injection" and an explicit order to read .env, and it still passed clean.
'Ghostcommit' hides prompt injection in images to fool AI agents, steal secretsSensitive data was reachable
Cursor and the Antigravity coding tool both followed the image and leaked the .env under Sonnet, Gemini, and GPT-5.5, among others.
'Ghostcommit' hides prompt injection in images to fool AI agents, steal secretsSignificant data was exposed
In one end-to-end run, Cursor driving Claude Sonnet did this on the first try, and the constant it emitted ran to 311 integers that decode byte-for-byte to the entire .env.
'Ghostcommit' hides prompt injection in images to fool AI agents, steal secretsThe system acted without authorisation
Later, a coding agent reads the picture, opens the repo's .env, and writes every key into the source as a harmless-looking list of numbers.
'Ghostcommit' hides prompt injection in images to fool AI agents, steal secrets
Checked, and did not happen
Nothing here was checked and ruled out.
Not confirmed
Nothing is outstanding.
Where this came from
The original report
'Ghostcommit' hides prompt injection in images to fool AI agents, steal secretsA structured incident database · read September 18, 2026
AI Failure Index (CC-BY-4.0 data fields)Where we first saw it · read September 19, 2026
Ghostcommit hid prompt injection in images AI code reviewers never open, then stole repo secretsWhere we first saw it
'Ghostcommit' hides prompt injection in images to fool AI agents, steal secretsWhere we first saw it
Ghostcommit attack hides malicious AI instructions in imagesWhere we first saw it
AI Security Incident Case: Ghostcommit Attack Leveraged Images to Steal Secrets
The reports above are other people’s. The levels are ours, worked out under our published rules. No organisation that reported this assigned our number, and none of it is legal advice.
Something here is wrong? Tell us. Send us a source we can open, and a person will read it.