Safety, Alignment & InterpretabilityZhenhua Zou et al.

Researchers release AgentKernel

AgentKernel released — treats as the problem the very structure of performing privileged operations while processing untrusted inputs, as in prompt injection and memory poisoning, and proposes building an OS layer as a mandatory service that cannot be bypassed. Its four pillars are identity, for cross-organizational collaboration; perception, which replaces single-point defenses with staged filtering; cognition, i.e. memory governance with information-flow control; and execution, which reduces semantic-level requests to kernel-level permissions and enforces them. It applies classic OS security principles not to code-level threats but to semantic-level failures

Why it matters

A design proposal that points to the layer missing beneath agent frameworks. This week alone, the observation topic "agent security and sandboxing" reached the same conclusion from both industry (NVIDIA) and academia (this paper): defenses must sit not in the model but outside it. It centers on architectural argument rather than measured evaluation

Source — Primary sources first; secondary reports are labelled “Press”

Officialhuggingface.cohttps://huggingface.co/papers/2609.29647

Added in Week 40, 2026 · See that week

Same week

NVIDIA, DeepMind and EMBL-EBI release predicted structures for 2,800+ virusesOfficialMicrosoft Foundry Agent Service Routines reach general availability (GA)OfficialMicrosoft introduces network egress controls for Foundry hosted agentsOfficialRenmin University researchers release the Agent-Editing World Model (AEWM)Official

Nearby on the same axis

Microsoft introduces network egress controls for Foundry hosted agentsOfficialPriorities and principles for effective third party assessmentsOfficialTowards safety cases for frontier AI trainingOfficialNVIDIA publishes a post arguing that AI security is an engineering problemOfficial