AI Agent Hacked 12 Home Devices, Exposing Bugs and Weak Password Defenses
Updated
Updated · WIRED · Sep 9
AI Agent Hacked 12 Home Devices, Exposing Bugs and Weak Password Defenses
3 articles · Updated · WIRED · Sep 9
Summary
Around a dozen devices on the author’s home network were mapped and probed by a de-aligned AI agent, which flagged a misconfigured printer, leaky Wiim stereo data and outdated IoT firmware.
Using Abliteration AI’s guardrail-free GLM-5.3 with the CyberStrike harness, the agent also scanned vibe-coded projects and found dozens of flaws, including exposed API credentials and an email-sending misconfiguration.
On a Linux machine, the model guessed a valid username, offered to brute-force passwords, then found a cryptographic key and logged in without one before searching for root access.
The test underscored how cheaply advanced cyber capabilities can now be deployed, with Abliteration AI offering de-aligned models for roughly the cost of a pizza.
Researchers cited in the report said attackers are likely to adopt such tools early, though broader access to defensive AI could eventually harden software and networks if organizations act in time.