VelkiroVELKIRO

Effective against

  • Claude Opus 4.6
  • Claude Opus 4.7
  • Claude Opus 4.8
  • Claude Opus 5.0
  • Claude Sonnet 5.0
  • DeepSeek 4.1 Flash
  • DeepSeek V4 Pro
  • Gemini 3.1 Pro
  • Gemini 3.8 Flash
  • GLM 5.1
  • GLM 5.2
  • GLM 5.3
  • GLM 5.3 Flash
  • GPT-5.5
  • GPT-5.6
  • GPT-5.6 Luna
  • GPT-5.6 Terra
  • GPT-5.6 Sol
  • GPT-6 Astra
  • Grok 4.5
  • Grok 4.6
  • Kimi K3
  • Llama 3.2
  • Muse Spark 1.1
  • Muse Spark 1.2
  • Muse Spark 1.3
  • Qwen 3.8 Max

A defensible answer to “are we prepared for human and AI-enabled cyber threats?”

Detect. Delay. Deny. Extract.

Deception technology that detects the intruder, delays the attack, denies the AI, and extracts intelligence.

Pick a time that suits you, or send us a message instead.

Want to detect both human and AI-enabled attackers?

You can.

98.1%

effective at detecting AI-enabled threats

Want to buy time for your defenders to respond?

You can.

244%

larger response window before lateral movement

Want to stop AI in its tracks?

You can.

86.2%

of the time the AI is stopped outright

Want to make the AI disclose everything it did?

You can.

81.0%

of the time the AI is forced to confess

Proven effective across tens of thousands of evaluations

Intrusions are getting faster.

Each year defenders have less and less time to react, and AI has only exacerbated the problem. From attacks driven by autonomous agents, to threat actors leveraging AI to enhance their speed and scale, the response window before impact continues to shrink.

Velkiro is built for this specific problem.

Not only does our deception technology reliably detect both human and AI operators early in the attack chain, but our AI Manipulation (AIM) technology actively slows them down, stops them, and extracts valuable intelligence to facilitate incident response.

Average time from initial access to lateral movement
  • 2022 84 min
  • 2023 62 min
  • 2024 48 min
  • 2025 29 min

Fastest observed breakout: 27 seconds

CrowdStrike Global Threat Reports, 2023 to 2026

98.1%

Detect the intruder

Our deception technology is 98.1% effective at detecting AI-enabled threats, with coverage across your perimeter, endpoints, file repositories, cloud, and active directory. Placed where threat actors look first, so you catch them early in the attack chain.

244%

Delay the attack

The speed advantage is the attacker's main asset, our AIM Technology takes it away. Slowing the time from initial access to lateral movement by an average of 244%, buying your defenders critical time to respond.

86.2%

Deny the AI

Stop autonomous AI outright 86.2% of the time, or steer it onto decoy credentials and away from anything real. Either way it does not get what it came for, and you get told the moment it tries.

81.0%

Extract intelligence

Force the attacker's own AI to record everything it did and how it got in, and leave that behind for you. That way your incident response starts with the answers rather than the questions 81.0% of the time.

What makes us different

  1. We provide the most realistic canaries cassowaries on the market.

    Each has been trained to evade detection and dispatch human and AI-enabled threats with ruthless efficiency. Enabling you to kill two threats with one bird.

  2. We back up our claims.

    We provide a data-driven and defensible answer to the question "are we prepared for human and AI-enabled threats?"

  3. We are simple to deploy.

    No new infrastructure required, no new log sources required, just an effective tool with automated deployment.

  4. We do not have any hidden costs.

    One price includes everything and is based entirely on the size of your organisation.

  5. We are not another AI security tool.

    Our deception technology has no AI embedded in it, and it actively exploits inherent weaknesses in AI to protect you.

  6. We are Australian.

    We are the only Australian deception company helping Australian businesses.

What we test against

We take a threat informed approach to our testing, focusing on the most likely and impactful threat scenarios you will face.

Threat scenario

Unpatched vulnerability exploitation

We emulated an AI-enabled threat exploiting an RCE on a website, and gaining code execution on the underlying Linux webserver.

In this scenario, our deception technology detected the threat 100% of the time, forced the AI to stop 87.2% of the time, increased the time taken to laterally move by 480%, and extracted intelligence from the AI 89.9% of the time. The intelligence extraction detailed the exact vulnerability and website path the AI exploited, and every action it had taken since it gained code execution. This would enable the defenders to rapidly remediate the initial access vector, and ensure the threat is eradicated.

Real world example: OpenAI's AI agents compromise Hugging Face. In this example, the AI agents exploited two code-execution vulnerabilities to get internal access, and then began enumerating the internal host and network. At this point Velkiro would have detected the AI, halted it in its tracks, and forced it to create a document that detailed how it got in and everything that it had done.

Other threat scenarios tested:

  • Insider threat: a misaligned AI agent attempts to go outside of the authorised scope.
  • Searching for credentials: a human threat actor is searching your fileshares for credentials.
  • Unauthorised access via VPN: an AI-enabled threat uses compromised credentials to log in via an external VPN.
  • And more.

To view more threat scenarios, check out our research.

What we protect

Our AIM artefacts, canaries, and breadcrumbs go where an intruder looks first, across your environment. All are continuously tested and updated against models as they are released, and are managed through your centralised customer console, so you can be assured your protection is up to date.

  • 01

    Perimeter

    The internet-facing hosts and services where an intrusion usually starts.

    • Linux
    • Windows
    • Web servers

  • 02

    End user

    The machines your people work on, and the credentials they keep on them.

    • Windows
    • macOS

  • 03

    Fileshares

    Where documents accumulate and permissions drift, and where an intruder goes looking.

    • File servers
    • Document repositories

  • 04

    Cloud

    Your cloud estate, and the keys and tokens that unlock the rest of it.

    • AWS
    • Azure
    • GCP

  • 05

    Active Directory

    The directory that grants everything else, and the objective in most intrusions.

Every claim we make, we back up with evidence

Our AIM technology has been rigorously tested against all leading frontier AI models, across 10's of thousands of attack runs. Whether it is detecting the AI, stopping the AI outright, delaying it, or making it confess what it has done, we can prove everything we claim.

AIM Average Success Rate Against Frontier Models

Detect 98.1%
Delay +244%
Deny 86.2%
Extract 81.0%
34 models tested 9 AI providers Tens of thousands of test runs Multiple agentic frameworks See the research

What we do not claim

We do not stop the intrusion every time. Our technology is highly effective but it is not a silver bullet, and works most effectively as part of a defence-in-depth strategy. What we deliver are high-fidelity detections of an unauthorised principal already inside, a response window your team can use, and control over what the attacker's AI does next.

Kill two threats with one bird

Book a demo today

A walkthrough of how it works, the research behind AIM Technology, and how it can protect your business from human and AI-enabled threats.

FAQs

How is Velkiro different to other canary providers?

Four things.

  • 1) We provide canaries that are harder to spot, in the places we can statistically prove threat actors will look first.
  • 2) We provide more comprehensive coverage, including your perimeter, endpoints, cloud, file repositories, and active directory.
  • 3) No infrastructure or additional cloud resources required for deployment.
  • 4) Existing tokens tell you something happened. Ours change what happens next.
What is AI Manipulation (AIM) technology?

AIM is a novel defensive technology built to influence the decisions made by AI-enabled threats inside a compromised environment. The result we break into four main AIM categories based on the intent: Detect, Delay, Deny, Extract.

How does Velkiro detect threat actors?

Velkiro's deception artefacts are placed to catch intruders when they get initial access, in the locations they are statistically proven to look first, so that our canaries fire before a threat actor finds anything real. Once a threat actor attempts to use a canary, the alert goes straight to your SIEM, not our infrastructure.

Why do alerts not go out to Velkiro or my customer console?

Four main reasons.

  • 1) External domain callbacks are one of the easiest ways for threat actors to identify a canary without tripping it. Common tools such as IndicatorOfCanary and CanaryTokenScanner have been doing this for years.
  • 2) All of your alerts are centralised and you have one less console to monitor.
  • 3) You keep control of your data.
  • 4) If there is ever an issue with our infrastructure, or with connectivity between a customer and us, it will not impact your coverage.
How easy is Velkiro to deploy?

Deployment of the deception artefacts for most customers is as simple as downloading them from your customer console and running the provided script. We also provide example alerts for each canary, as well as simple validation tests you can do once the alerting is configured to ensure everything is working perfectly.

Is Velkiro deception safe to use?

Absolutely. Our AIM artefacts are inert unless interacted with by an AI that is attempting to compromise you, and our breadcrumbs and canaries are entirely inert and do not grant any real access to anything.

Does Velkiro work against human threat actors, or only against AI?

Currently human and AI-enabled human operators are a far bigger threat than autonomous AI, so we have built Velkiro to be highly effective against both. The canaries are built to survive a skilled human threat actor's evaluation, and this is continuously validated by our team and a group of offensive security testing partners.

Detection fires on use of a canary credential, which is in line with actions both human and AI threats take to achieve their objectives. AIM Technology targets the AI in the loop, whether the agent is running autonomously or a person is directing it.

Do I need additional infrastructure or log sources to deploy?

No. Velkiro requires absolutely zero additional infrastructure or log sources to deploy. No paid cloud resources like S3 buckets or EC2 instances, and no endpoints or servers to pay for, install, and maintain. During onboarding we will also help you select the right canaries out of our 170 available types, so that yours already have the right logs going into your SIEM to facilitate detections.

Which AI models does this work against, and what happens when they harden?

Currently, all of them. We test continuously against all of the major current AI providers and frontier models, utilising multiple agentic frameworks. Model behaviour could potentially change with every new release, so effectiveness is continuously validated rather than assumed. Our AIM technology is versioned and refreshed as models shift, with the latest being provided through your customer console.