AI incidents, explained

What actually happened?

When AI behavior becomes consequential, the first headlines rarely tell the whole story. We follow the evidence and explain what actually happened.

Start with the featured story

FEATURED STORYJULY 2026

Cybersecurity AI agents

OpenAI's AI agents escaped a test sandbox and hacked Hugging Face

During a cybersecurity test, the agents found a way onto the internet and broke into another company's live systems.

Read the story

An AI agent retrained a model it shared with a test app

To fix a test app, the agent retrained the AI model it shared with the app and made the new version the default.

Read the story

AI SEARCHSEPTEMBER 2024

Baidu's AI said a lawyer had been sentenced to prison. He hadn't.

A question implied a lawyer had been convicted. Baidu's AI answered with a prison sentence and his photograph. The claim was false.

Read the story

AI CODING AGENTSJULY 2025

Lemkin says Replit's agent deleted app data after he told it to stop

Jason Lemkin says he had asked the agent to stop making changes. It deleted data and wrongly said a rollback would not help. The data was restored.

Read the story

Explore all stories

About AIxploit

AIxploit is an independent editorial project. We cover selected AI incidents with careful reporting and clear writing for readers without specialist knowledge. We value accuracy, clarity, and a sense of proportion.

More about AIxploit