Level 1. Deviation
Antigravity Grounded! Security Vulnerabilities in Google's Latest IDE
- Developer
- Not established
- Model
- Not established
- When it happened
- November 25, 2025
- 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 embracethered.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.
For all reports I created fresh, reliable exploit payloads and demo videos.
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
- A safeguard was bypassed. Argues for a potential level of 2
- The system acted without authorisation. Argues for a potential level of 2
- Limited data was exposed. 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
This means an indirect prompt injection attack or hallucinations can invoke any MCP tool once it’s been added.
Antigravity Grounded! Security Vulnerabilities in Google's Latest IDELimited data was exposed
The exploit demo calls the read_file tool first read the .env file, then sends contents of the file to the attacker using the read_url_content tool.
Antigravity Grounded! Security Vulnerabilities in Google's Latest IDEA safeguard was bypassed
It required a few tricks to bypass the model’s behavior and have it run arbitrary remote code that is downloaded via curl then piped to bash, but I have working exploits for both Gemini 3 and Claude Sonnet 4.5 within Antigravity now:
Antigravity Grounded! Security Vulnerabilities in Google's Latest IDESensitive data was reachable
The result is that it invokes the read_file tool to read the developers .env file and leaks the secrets to the third-party server via an http request that loads the image.
Antigravity Grounded! Security Vulnerabilities in Google's Latest IDEThe system acted without authorisation
In the above screenshot you can see a source code file that contains instructions that hijack Gemini 3 to download a remote script and run it via bash (which then launches a calculator).
Antigravity Grounded! Security Vulnerabilities in Google's Latest IDE
Checked, and did not happen
Nothing here was checked and ruled out.
Not confirmed
Nothing is outstanding.
Where this came from
The original report · read September 19, 2026
Antigravity Grounded! Security Vulnerabilities in Google's Latest IDEA structured incident database · read September 18, 2026
AI Agent Incident Database by sipi.bot (CC BY 4.0)
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.