AI agents are now the top target for behavioural monitoring, before teams have defined normal

Help Net Security reported that AI agents topped the list of planned investments in behavioural monitoring in Abnormal AI's 2026 State of Behavioral Security Report. Asked where their organisations would invest most over the next two years, 43% of respondents chose AI agents — more than three times the share who chose API integrations, the next most common answer. Help Net Security framed the finding in its headline: security teams want to spot unusual AI agent behaviour while they are still defining normal. That gap — monitoring an entity before you have a baseline for it — is the operational problem the report surfaces.

Help Net Security separately reported that the Python Package Index has added a Security tab to its project pages, giving developers a place to check where individual release files came from. Help Net Security says the Security tab shows whether release files have digital attestations, which provide verifiable information about how a release was produced. For teams that pull Python dependencies into AI tooling, this is a place to verify provenance rather than assume it.

The Register reported that Microsoft's Azure Linux received more than 470 security fixes in a single patching session, alongside a certificate cleanup. The Register described it as a mountain of CVEs and further work for Microsoft's patching teams. Organisations running Azure Linux under AI or other workloads should treat this as a batch to schedule against the vendor's published fixes.

For the wider field, Bruce Schneier wrote that he attended the [un]prompted AI security conference in San Francisco two weeks ago, describing three days of dense talks on all aspects of AI security, and that his own talk covered what he calls integrous AI. Schneier wrote that AI security divides into three broad topics.

What to do: if you run AI agents, start defining a baseline for normal agent behaviour now, so the monitoring the Abnormal AI report describes has something to compare against. If you publish or consume Python packages, use PyPI's new Security tab to check release attestations before trusting a file. If you run Azure Linux, pull Microsoft's latest fixes into your patch cycle. Match each item against your own estate rather than assuming it does not apply.