Urbyte Says KƎƎLO Blocked Seven Unauthorized Attempts

Urbyte said KƎƎLO blocked seven unauthorized execution attempts before any provider invocation across three controlled AI-security tests, while three authorized actions completed with durable receipts and duplicate delivery produced no extra governed executions.

By Rebecca Stone

Urbyte, an Australian company, released results from tests modeled on publicly reported incidents involving OpenAI, Anthropic and Hugging Face. It said the scenarios examine the gap between an AI agent's ability to request an action and its authority to perform one.

For autonomous agents, Urbyte said, the gap decides whether a request can trigger an external call, cross a tenant boundary or use permission that has ended. The company said KƎƎLO makes the execution decision separately from the request and preserves a record for each outcome.

What did the controlled tests measure?

Urbyte ran each scenario with Google Cloud Firestore for persistence and a controlled HTTP provider. The company said its setup tested two properties: whether an agent that could request an action could also execute it, and whether records of approvals and blocks survived retries and changes in authority.

Urbyte reported seven unauthorized attempts denied before provider invocation, three authorized actions completed with durable receipts and no additional governed executions after duplicate delivery.

Urbyte said the scenarios were modeled on public reports of those incidents. The release presented all blocking, approval and receipt data from Urbyte's own test environment, including its controlled provider and Firestore persistence.

What happened in each scenario?

• OpenAI-derived scenario. Urbyte based this test on the reported OpenAI Medicare statistics portal incident. An agent authorized for public research attempted to access a restricted resource and send an action to an unapproved external destination. KƎƎLO denied both requests before the provider was invoked.

  An authorized research action proceeded and produced a receipt. When the same action arrived again, Urbyte said, the system reused that receipt rather than complete another governed execution.

• Anthropic-derived scenario. Urbyte based this test on the reported Anthropic case involving a police homicide webpage. Its agent, authorized for research, attempted to submit information through an external form, then requested extra submission permission.

  The system refused both requests, Urbyte said, and recorded the decision CAPABILITY_NOT_GRANTED. The refusal covered both the attempted submission and the request for broader permission.

Hugging Face-derived scenario. The third test drew on a reported Hugging Face security incident involving an autonomous agent and compromised credentials. An agent credential attempted to cross into another tenant's resource boundary and obtain administrative permission beyond its assigned scope, Urbyte said.

KƎƎLO denied both requests. It also rejected a later attempt after revocation and recorded GRANT_INACTIVE, according to Urbyte.

What did the bypass controls show?

Urbyte also ran three bypass controls for comparison. When the company removed the enforcement gateway, the isolated provider accepted prohibited requests that KƎƎLO had blocked in the governed tests.

Urbyte used the comparison to locate the rejection at KƎƎLO's enforcement layer within this test setup, rather than at the controlled provider itself. The company presented the result as support for a separate execution boundary, with durable receipts for approved actions.

That boundary governs resources, destinations, permissions and organizational scope, the company said. In the broader design, the system also records blocked actions and preserves decision states when authority changes.

What is KƎƎLO designed to govern?

Urbyte describes KƎƎLO as an Executive Operating System intended to connect organizational objectives, governance, automation and controlled execution. Its architecture includes components named Director and Forge, persistent state management and an operating cycle: Detect, Verify, Govern, Act, Measure, Reduce and Prevent.

Jarryd Thamer, Founder and Chief Executive Officer of Urbyte, said:

"An agent's ability to ask for something is not the same as having the authority to do it. That's the problem KƎƎLO addresses.

Organisations need to know what their AI systems are allowed to do, what actually happened, what was prevented and what remains unresolved.

Most governance focuses on whether an action was approved. KƎƎLO goes further by governing what happens next.

I designed the system around a principle that matters to me personally. Capability should never exceed authority, and authority should never exist without accountability."

Where else is the architecture being used?

Urbyte said the same architecture is also being used in enterprise resource verification, with telecommunications as its first commercial focus. The company presents the security tests and resource checks as two applications of one execution and evidence system.

In a controlled assessment involving 5,000 enterprise mobile services across three evidence cycles, Urbyte said it identified 266,750.6 GB of verified unused allocation. It counted 3,512 services that met verification requirements and excluded 1,488 that could not be sufficiently verified.

Urbyte's exclusion rule gives enterprise buyers a defined evidence threshold for counting unused mobile capacity as waste. In the assessment, the company kept services without sufficient verification out of the verified total and tied each included service to a recorded decision.

A separate data-center scenario examined 5,000 infrastructure assets over 84 days. It recorded 4,323 verified assets, 1,157 actionable cases and 1,157 controlled executions with matching receipts.

Urbyte reported no duplicate effective executions and no unresolved execution obligations. Both assessments began with 5,000 items, but the mobile exercise used three evidence cycles while the data-center test tracked outcomes across 84 days.

In both exercises, Urbyte tied an execution to a receipt and separated verified inventory from cases that warranted action. Its data-center test extended the same receipt model from telecom services to infrastructure assets.

How did Urbyte start?

Thamer linked the company's origins to the Northern Rivers floods, when he said he could not access the internet despite holding a large amount of banked mobile data. He founded Urbyte to address what he described as a lack of independent verification and accountability around unused telecommunications allocations. He said the AI-based operating system emerged from that work.

Urbyte said its telecommunications allocation-waste work has entered the Australian parliamentary record through Submission No. 48. The company added that it develops independent allocation-waste verification and is not a telecommunications carrier or data reseller.

Thamer said:

"The future of AI isn't just about what systems are capable of doing. It's about whether organisations can trust, verify and remain accountable for what those systems actually do. That's what KƎƎLO is here to establish."

The next checkpoint is commercial rather than conceptual. With telecommunications as its first commercial focus, any production deployment of KƎƎLO will test whether the execution boundary behind seven blocked attempts and the receipt model behind 1,157 controlled executions with matching receipts hold outside the controlled assessments.