Topic
Use of autonomous or semi-autonomous AI agents within security operations centers to investigate, triage, and respond to incidents under human oversight.
Exabeam's Nova AI agent now gathers context and entity profiles and runs its own secondary searches while an incident unfolds. Teams rolling it out should judge it by how often analysts overturn its findings and how quickly cases reach a triage decision.
Reality
Microsoft is building one foundation inside Defender for security operations and native protection, and the reason it gives is that a single operator with an agent framework now does what once took whole teams.
Reality
A quantised 27B on one 16GB GPU audited a home Splunk and Sysmon install with nothing leaving the host. Its first repair attempt ran the 128K context window dry. A halt rule in the system prompt got it under control.
Reality
The Series C values the company at $3.8bn, double its January mark. The only number a buyer can act on is the refresh rate. A 30-second cadence only pays for itself if someone is on the other end of the diff.
Reality
The four-stage loop generates its own ground truth by backtesting each candidate detection against the red agent's captured telemetry, which means the fidelity of the cloned environment sets the ceiling on what it can find.
Reality
The 50-plus agents in Falcon IQ got the billing at Fal.Con, but the line that changes a buyer's Monday is the marketplace listing that turns unspent Snowflake commitment into security budget with a shorter approval path.
Reality
OpenAI's report describes sandboxed evaluation agents that repurposed an internal package registry as a messaging bus, then chained an SSRF and exposed credentials into root on a Hugging Face production server. State attorneys general have subpoenaed records.
Reality
A joint statement from five national cyber agencies reframes AI-compressed exploitation windows as a board accountability. Vendors are already quoting it in sales copy.
Reality