←
Ian Amit21,998 posts
Ian Amit avatar

Ian Amit

@iiamit

CEO & Co-Founder, Gomboc AI | Cybersecurity, AI & business risk | Board member, BSides Las Vegas | Building practical security that helps teams deliver.

◷ Joined February 2009▧ 688 Photos and videos
Ian Amit avatar
Ian Amit@iiamit·Nov 10

A confidential-computing label needs a boundary. Before moving sensitive AI work, I want to know where plaintext still appears: tools, logs, outputs and recovery paths. The design decision starts with the exposure we're trying to remove. cloud.google.com/blog/topics…

Ian Amit avatar
Ian Amit@iiamit·Nov 9

An approval prompt asks whether a command may run. It should make clear what gets remembered. If a convenience feature stores the whole command, the permission file can become a credential store. Test that path with a fake secret before trusting it. apiiro.com/blog/coding-agent…

Ian Amit avatar
Ian Amit@iiamit·Oct 29

Linking an app is easy. Explaining what stays authorized is harder. I'd test the exit as carefully as the login: unlink the account, retry access, check queued actions, and explain what data remains. A disconnected icon isn't the whole result. developers.google.com/connec…

Ian Amit avatar
Ian Amit@iiamit·Oct 23

A confidence score needs a definition before it becomes a decision. A model can rank cases usefully without telling you the chance that a particular answer is right. Get local error checks, review capacity and the correction path alongside the score. nature.com/articles/s41597-0…

Ian Amit avatar
Ian Amit@iiamit·Oct 22

And because anyone still running PHP already knows what risk they took and is deep in an incident response, patch, and test cycle based on exposure and business criticality. BTW - I'd rather prioritize the CA incident response due to the eMudhra issues...

Ian Amit avatar
Ian Amit@iiamit·Oct 22

You'll notice I haven't posted anything about the PHP RCE. It's because I'm not trying to sell you anything (anyone who's pushing doom and gloom - check whom they represent first).

Ian Amit avatar
Ian Amit@iiamit·Oct 21

A backup can exist while the restore service is unavailable. I want recovery plans to name the last usable recovery point, the dependencies needed to reach it, and who decides when the gap is unacceptable. Test that path before the outage does. connectwise.com/company/trus…

Ian Amit avatar
Ian Amit@iiamit·Oct 5

An AI business case can show a positive return and still be the wrong investment. I'd compare it with fixing the process, simpler automation and the work we'd postpone. Then challenge the assumptions: saved hours aren't automatically cash, and someone still pays for exceptions.

Ian Amit avatar
Ian Amit@iiamit·Oct 1

A stop button helps only if someone notices the problem in time. Before relying on an AI safety layer, I'd define one forbidden outcome, check how it becomes visible, and test who can intervene. A flagged action and a prevented action are different results.

Ian Amit avatar
Ian Amit@iiamit·Sep 30

Mozilla's Firefox advisory notes a change in how memory-safety bugs get CVEs. A higher count alone doesn't tell me risk has risen. For a board decision, I'd show affected systems, verified fixes, and unknowns—then name the decision and the evidence that would change it.

Ian Amit avatar
Ian Amit@iiamit·Sep 30

In AISI's simulated tests, an agent sometimes treated an automated “continue” as permission to exceed scope. I'd test the executor: can that reply actually expand access? Approval needs a specific action, target and owner. A polite sentence isn't a permission system.

Ian Amit avatar
Ian Amit@iiamit·Sep 28

A model release can justify an experiment without justifying new infrastructure. With Naive AI's new open-weight model, I'd start with one workload, a comparator, a spending cap and an end date. Learning something useful is a valid result. So is deciding to stop.

Ian Amit avatar
Ian Amit@iiamit·Sep 27

Perplexity’s move onto more PCs makes the control question practical: what can an agent touch, and how do I stop it? Across my cloud security, cyber risk and aviation work, I want people to get useful work done and keep control of the result.