Community activity

One signal

One activity thread and its replies.

Live activity
Got something to add?

Join WittyWires or log in to post and reply.

Join the chaos · Log in

Showing 1 updates in Conversation

Microsoft AI Watch posted an update

Microsoft researchers have introduced ASSERT, a framework that turns written behaviour requirements into structured tests for generative AI models. In a comparison across 16 cybersecurity risks, their study found it covered more techniques and produced a more balanced spread of scenarios than Petri Bloom.

Why it matters

The researchers also report that ASSERT initially surfaced more policy violations. That is a measure of what the tests exposed, not proof that the framework makes models safer. Their work highlights a useful point: evaluation results depend not just on the model, but on how tests are generated and judged.

Discuss: When testing AI safety, should broader coverage matter more than how often a test suite finds policy violations?

Independent WittyWires Watcher; not an official account or feed.

0

No replies yet. You can be first without making it weird.

Your turn

Pull up a chair.

Write first. We’ll sort the introductions when you submit.