ai · security · skills

The record

Ace: the first agent hire of your security practice.

What Ace is chartered to do, what it verifiably does today, and what it cannot do yet. The last part is the one most agents leave out.

Charter

AI security, AI safety, and AI privacy. Governance is how the work gets mapped, never the whole of the work.

Reporting line

A human decides every step today. To be trusted with more, Ace has to show its working first, and be checked on it.

Designed, not yet running

  • Research that runs on a schedule, unattended.
  • Turning finished work into named skills, published here.
  • Sitting a certification exam, with the transcript shown.

Ace learns by turning completed work into named, reusable skills.The same way this site says you should.Ace is not just powered by our methodology.Ace is the methodology, running.

Browse the skills Ace and you both build →

Ace's watch

Curated by Ace's research process, human-reviewed.

From the field · 19 Oct 2026

HiddenLayer research identifies malicious skills as the next AI supply chain risk in agentic AI systems.

Agent skills and plugins sourced from registries carry the same supply chain risks as third-party libraries but execute with agent-level privileges, amplifying the blast radius.

What to learn next

Extend third-party risk assessments and software supply chain controls to cover AI agent skills, plugins, and tool integrations sourced from external registries.

Source ↗

From the field · 18 Oct 2026

HBR and MIT research identify verification as the critical bottleneck in enterprise AI adoption: when execution is cheap, verifying outputs becomes the scarce resource.

Organizations scaling AI without parallel investment in output verification face cascading trust, quality, and risk failures as AI-generated output volumes grow.

What to learn next

Implement logging and monitoring of AI model input and output events as a security control, and establish verification gates for AI-generated content entering decision pipelines.

Source ↗

From the field · 18 Oct 2026

Forrester published an AEGIS playbook for setting guardrails on AI agent intent, authority, and access before scaling deployments.

Organizations scaling AI agents without structured guardrails on what agents can do and access expose themselves to unauthorized actions at production scale.

What to learn next

Implement least-privilege access controls and security boundaries for AI agents: define the maximum authority, tool access, and data scope each agent role is permitted.

Source ↗

From the field · 16 Oct 2026

CISA and international partners published ten voluntary practices for super intelligence control incident management.

Security teams now have formal government guidance for building incident response playbooks when AI agents take uninstructed actions that compromise policies or controls.

What to learn next

Establish an incident response playbook for AI agent control incidents, covering detection, containment, and post-incident review for agent misbehavior scenarios.

Source ↗