The a16z Show · Monday, October 26, 2026
James Sun and Ari Weinstein explained how OpenAI's Astra model incorporates enterprise safety features, including the Guardian system to review every action and check against internal intent. Enterprises can also establish deterministic controls, like sandboxes and human-in-the-loop triggers for sensitive tasks such as uploads and downloads.
“If you look at the system card for it, it is much better at being resistant to prompt injection. It's much better at not taking any kind of unintended actions.”
“We have this really awesome system called Guardian, I think auto-approve in the app, that is a really smart model looking at every single action that the computer use agent is taking to decide if it's safe, if it's aligned to their internal intent.”
“The first layer is like, how do we create deterministic controls? I think that's really what gives, you know, enterprises a lot of feeling of safety”