Silence Laboratories wins Accenture Tech Next Challenge 2026

Silence Laboratories has been named the Cybersecurity Track Winner at the Accenture Tech Next Challenge 2026.

At the finals in Bengaluru, we presented AP3, our Agent Privacy-Preserving Protocol for secure collaboration between AI agents. The recognition validates a principle central to our work: as agents become part of critical enterprise workflows, privacy and trust must be built into the infrastructure through which they communicate.

The privacy problem in multi-agent systems

AI agents are beginning to move beyond isolated tasks within a single organisation. They are communicating across departments, companies, vendors and jurisdictions to complete increasingly complex workflows.

To work effectively, these agents need context. That may include customer information, financial records, credentials, risk signals, proprietary policies or commercially sensitive data.

Today, enterprises primarily protect this information through access controls and redaction. These measures can restrict what an agent sees, but they also restrict what it can do. Removing too much context makes the agent less useful. Providing more context increases the risk of exposing raw data across organisational boundaries.

This creates a fundamental trade-off: the more an agent knows, the more sensitive information it may reveal.

AP3 is designed to break that trade-off.

Enabling collaboration without exposure

AP3 allows agents to collaborate and compute over sensitive information without exposing their underlying private inputs.

Two agents that do not trust each other can use secure multi-party computation to generate a joint result while keeping their respective data inside their own environments. Neither agent sees the other party’s raw information. Only the result agreed upon by both parties is revealed.

The exchange is also tied to each agent’s cryptographic identity. Messages and contributions are signed using the agent’s private key, making interactions attributable, tamper-evident and auditable by default.

This creates a foundation for collective intelligence without requiring organisations to surrender control of their data.

How AP3 works

AP3 introduces privacy and policy enforcement directly into the multi-agent workflow.

Developer-friendly private computation

Privacy-preserving computation has traditionally required developers to write logic using specialised or restricted languages.

With AP3, developers can express business logic using familiar programming languages. This logic can then be compiled into a privacy-preserving and auditable circuit that agents can evaluate securely.

The result is a more practical path for bringing advanced cryptography into existing enterprise applications.

Independent policy checks

Before a computation begins, each participating organisation can evaluate the proposed circuit against its own disclosure and compliance rules through the AP3 policy engine.

Both parties must approve the computation before it can run. This ensures that no agent is required to participate in a workflow that violates its organisation’s policies.

Only the specific result authorised by both parties is allowed to cross the trust boundary.

Computation without data movement

Once approved, the agents compute jointly using secure multi-party computation.

Private data is divided into cryptographic shares and processed collaboratively. The original information remains within the environment of the agent that owns it. No participant, intermediary or infrastructure provider gains access to the complete underlying data.

Compatibility with agentic commerce

Once a result has been approved, the workflow may need to proceed into a commercial action such as payment or settlement.

AP3 supports AP2 and x402, allowing commerce to continue within the same agentic flow without requiring a separate integration layer.

It is also framework-agnostic by design. AP3 can work with agentic frameworks including Google ADK, LangGraph, AutoGen and CrewAI. Built on A2A rails, it is designed to support private agent communication across providers and enterprise environments.

What does this enable?

AP3 can support workflows across finance, healthcare, commerce and other industries where collaboration is valuable but unrestricted data sharing is not possible.

Transaction monitoring

One agent can check a transaction against another organisation’s fraud or sanctions data. The requesting agent receives the required result, such as a yes or no response, while the underlying watchlist remains private.

Healthcare collaboration

Hospitals and healthcare providers can identify patient overlap or evaluate shared risk signals without exchanging entire patient databases.

Cross-institution anti-fraud checks

Banks can identify shared bad actors, suspicious accounts or coordinated fraud patterns without pooling customer information into a central database.

Private underwriting

Business and lending agents can collaborate on sensitive financial information to assess creditworthiness and generate an underwriting decision.

Multiple lenders can evaluate an application and return their proposed terms without receiving unrestricted access to the applicant’s financial records. The underlying data is disclosed only when required by the final workflow.

Building privacy into the agentic economy

As agents begin making decisions and coordinating across organisational boundaries, access control alone will not be enough.

Enterprises will need infrastructure that can verify who participated in a workflow, control what was computed, enforce each party’s policies and prove that sensitive information was not unnecessarily exposed.

Privacy cannot remain an additional layer applied after an agentic system has been designed. It must become part of the protocol through which agents discover, communicate and collaborate.

Winning the Cybersecurity track at the Accenture Tech Next Challenge 2026 is an important validation of this direction and of the work being done by the entire Silence Laboratories team.

Our thanks to Accenture and the jury for the recognition.