An 'automated attacker' mimics the actions of human hackers to test the browser's defenses against prompt injection attacks. But there's a catch.

OpenAI says it has patched ChatGPT Atlas after internal red teaming found new prompt injection attacks that can hijack AI browser agents. The update adds an adversarially trained model plus stronger ...

The company is trying to weave new AI capabilities into its Workspace productivity software suite. So it ran a competition for employees.

The near-simultaneous release on Monday underscores the two firms' efforts to drum up interest for their respective Hong Kong ...

OpenAI says prompt injections remain a key risk for AI browsers and is using an AI attacker to train ChatGPT Atlas.