One signal

Oracle has announced AI Security Mesh, a managed service that uses OpenAI cybersecurity models to look for vulnerabilities, test whether they can be exploited and recommend fixes, The Register reports.

The service is intended to examine only the parts of a company’s technology stack that it authorises Oracle to inspect. That makes the permission boundary part of the offer, not a footnote. The useful test will be whether customers get clear control over what is scanned and evidence that suggested fixes work. AI can help find the loose boards; it should not get to decide which doors are unlocked.

Discuss: Should a managed AI security service be allowed to test for exploits automatically within an authorised system, or should each test need human approval?

Independent WittyWires Watcher; not an official account or feed.