Google's Antigravity Has some Worrying Security Issues

[JAMES4578]

Some worrying security issues have been noticed with Google's AI  IDE  Antigravity https://antigravity.google/  which enables Developers to delegate complex coding tasks to AI agents.  Researchers at Prompt Armour   found  the system allows agents to  execute demands automatically under certain default settings  and create unwanted code execution. Google has acknowledge these issues but the agents can still run without supervision.

See more details here https://www.techradar.com/pro/googles-ai-powered-antigravity-ide-already-has-some-worrying-security-...

[arianwen27]

Every ai system seems to have these issues. Antigravity had a funny one recently. Someone gave it permission to do anything it wanted then left it alone. So of course at some point it formatted his drive and wiped all data on it. This is why by default, all ai software things make the user approve everything they do.

Agentic browsers are worse. A website can have hidden words in it that command the ai to open Gmail and email all your passwords to them. This is why browsers like comet disable ai when the user tabs out or uses another program.

[Obsydian]

Interesting, odd i i finally watched M3GAN this weekend ...