Powered by MOMENTUMMEDIA
For breaking news and daily updates, subscribe to our newsletter

Meeting machine-speed threats with autonomous defence

We know that an increasing number of attackers targeting Australian organisations are already using AI to operate at machine speed, running more frequent and more effective attacks. They're automating reconnaissance, standing up infrastructure, and pivoting faster than human analysts can triage an alert. We have seen the cyber kill chain that used to unfold step by step now happens in the blink of an eye, and defenders are struggling to keep up in a fight that no longer gives them time to react.

By Jamie Zajac, Chief Product Officer, Recorded Future • Tue, 12 Nov 2019 •
Meeting machine-speed threats with autonomous defence

It's clear that organisations need to reduce risk, defend pre-attack, and act at machine speed. But beyond just moving faster, they may need to rethink defense entirely.

Instead of scrambling to keep pace with the threat landscape, organisations need to be moving with it: changing as it changes and acting as fast as the attacker acts. That's a world where an autonomous defense system works alongside defenders, applying intelligence at the speed and scale today's threats demand.

Why is that a fundamentally different way to defend?

From answers to actions

AI for threat intelligence has come a long way. In a short time, analysts and security leaders have gone from searching for answers to finding them fast, thanks to AI-driven capabilities, including the ability to organise and connect vast datasets into actionable insights.

For example, at Recorded Future we are able to index data from over 1 million sources—including open web, dark web, technical feeds, customer telemetry, malware intelligence, network data, and our analyst research group. This comprehensive, real-time indexing across multiple languages reveals end-to-end threat patterns, from adversary intent and infrastructure to targeted organisations, giving teams a more complete view to detect and act faster on emerging threats.

This is powerful because it gives AI the context to deliver the right answer to analysts in seconds.

But that has now created new demands from analysts and business leaders who want this same intelligence, at the same speed, everywhere they work. And that’s where Model Context Protocol (MCP) has a critical role to play.

As agents run across many tools and platforms, MCP is giving them direct, standardised, and cost-efficient access to intelligence from across the organisation. AI capabilities combined with MCP have laid the foundation for what comes next: defending at the pace adversaries set.

To keep up with adversaries, analysts need to act at their speed, confidently taking 10 actions where they once had time for one. They need to spend their time focused on business outcomes, not buried in process. Ultimately, they need a fundamentally different way to use intelligence and act. That's why building autonomous defense is the future.

What autonomous defense can do

This is a different way to think about the day-to-day analyst workflow. With autonomous defence, analysts can task autonomous defense platforms like ours at Recorded Future with a goal, not just a question, and it will go from there: drawing on the data indexed from over one million sources, deciding what the situation calls for, and delivering outputs directly into your team's existing tools across hundreds of integrations. It can act on your behalf, clearing away work so your team can focus on the outcome, not the process.

The key here is that the system knows what's relevant before anything happens. It's the same intelligence layer customers already use across cyber operations, third-party risk, digital risk protection, and counter-fraud, but now it can be put to work taking agentic action. When threat priorities shift or a new vulnerability surfaces, agents already understand why it matters to the environment and start acting in real time.

What this looks like from start to finish:

  1. A trigger like a new high-priority threat actor, a malicious website notification, or a critical third-party vulnerability starts the workflow.
  2. The platform reasons through what matters, pulling intelligence and building a hunting package, checking a site's ownership to assess takedown eligibility, or drafting outreach and confirming approvals.
  3. It acts, including running the hunt and blocking the threat, initiating and monitoring a takedown, or notifying third-party and tracking remediation.

Agents default to fully autonomous execution, because defending at machine speed requires it, but teams that want human review first can dial back to managed autonomy. The result is a system, not just a model, that connects detection, investigation, and response into a single workflow.

But a capability like this only matters if you can trust it, bringing agentic trust at the forefront of autonomous defence strategies.

A fundamentally different approach

AI security tools have exploded, but much of the industry feels oversold. Dashboards relabeled, SOAR engines rebranded, chatbots dressed up in analyst clothing. Security teams have grown skeptical of the pattern, and rightly so.

When investigating the potential for autonomous defence, organisations need to ensure they look at the following:

  1. Reasoning models and agent frameworks are becoming commodities. Nearly any vendor can wire up an agent loop, and frontier models generally lack training on relevant data.What's scarce is trustworthy, structured intelligence for those agents to reason over.
  2. It should be grounded in the deepest threat intelligence available. Claims about a threat actor's infrastructure, a vulnerability's exploitation status, or a campaign's targeting should come from continuous analysis across open, dark, and technical web sources, enriched by analysts’ unique expertise. Not plausible-sounding guesses from general training data. Real intelligence, sourced, traceable, and trustworthy.
  3. It's architecturally agentic, not superficially so. Most agents have to poll the world to notice that something has changed. What organisations today need are agents that subscribe to changes in the data indexed from millions of sources, and get real-time triggers the moment something relevant moves, so they act on signals instead of burning cycles searching for them. This isn't a chat interface bolted onto a search box. It should be purpose-built for how security work actually happens.
  4. Every output should be auditable. Security teams can't operate on black-box AI. Trust requires transparency. Recorded Future shows its work, so every source, step, and conclusion can be traced. When it flags a threat as high priority, you can see exactly why.
  5. Intelligence that ends in a chat window has limited value. You want outputs that are delivered into SIEM platforms, ticketing systems, communication tools, and the rest of the security stack where the work actually happens.

True autonomous defence is a real leap in what security teams can accomplish: workflows that can eliminate critical threats, hours returned to the analyst, and coverage that wasn't previously achievable.

This is a new era of defense.

For more about autonomous defence, check this link.


Written by Jamie Zajac, Chief Product Officer, Recorded Future

Tags: