Red-Teaming the AI Workforce: Workday and Straiker End the 'Trust Us' Era

MOUNTAIN VIEW, Calif. – October 16, 2026 – For years, enterprise software procurement has operated on a polite fiction: vendors promise their applications are secure, and buyers largely take their word for it, backed by static compliance checklists. But as autonomous AI agents begin accessing highly sensitive corporate workflows—from processing payroll to onboarding employees—that paradigm is collapsing. The risk of an AI agent hallucinating a financial transaction or falling victim to a malicious prompt injection is too severe for self-attestation.

Today, Straiker, a cybersecurity firm specializing in agentic vulnerabilities, announced a pivotal integration with Workday's Agent Passport. The partnership connects Straiker's Ascend AI red-teaming platform directly to the enterprise software giant's Agent System of Record (ASOR). The goal is as ambitious as it is necessary: establishing the first commercial, cryptographically verified audit standard for AI agents operating in high-stakes corporate environments.

By replacing vague vendor assurances with independent, adversarial testing, this integration signals a broader industry shift. Organizations are demanding concrete evidence of security before allowing autonomous logic to touch their most critical data.

The Enterprise Trust Layer: Moving Beyond Vendor Promises

Unveiled earlier this year at Workday DevCon and slated for general availability by the end of 2026, Agent Passport functions as a comprehensive trust layer for the burgeoning AI ecosystem. It is built directly into the ASOR, a centralized hub that allows IT and business leaders to register, configure, and monitor every AI agent operating within their infrastructure.

Agent Passport effectively breaks down trust into three distinct layers: broad categories defined by Workday (such as runtime safety and human oversight), specific testable claims tied to public standards, and cryptographically signed results from independent testing partners.

These signed results, referred to as "Stamps," are visible to customers before an agent is ever installed. If an enterprise requires an agent to possess a Stamp for "Prompt Injection Resistance," and the agent lacks it, the system can automatically block its deployment.

“As AI agents take on real work in HR and finance, ‘trust us’ is not a governance strategy. Customers need to know exactly what has been verified about an agent, by whom, and to what standard,” said Dean Arnold, VP of Workday's AI Platform. “Straiker brings rigorous, continuous adversarial testing to that question, and making their results a first-class input to Agent Passport will give our customers evidence they can act on.”

Crucially, this system introduces the concept of "Agent Owners"—assigning a named, accountable human to every autonomous application. When combined with real-time monitoring that can instantly revoke an agent's access if its behavior deviates from established parameters, organizations are finally gaining the granular control required to manage a non-human workforce.

Red-Teaming the Autonomous Workforce

Securing an AI agent requires fundamentally different tactics than securing traditional software. Traditional applications follow deterministic logic; AI agents operate probabilistically, making them susceptible to novel attack vectors like prompt injection, where malicious instructions are hidden within seemingly benign inputs to hijack the agent's core functions.

Straiker's Ascend AI platform addresses these unique vulnerabilities through continuous adversarial testing, commonly known as red-teaming. As the first independent security partner to evaluate customer- and partner-built agents for the Workday Marketplace, the cybersecurity firm probes these applications against established industry frameworks, including the OWASP LLM Top 10, the OWASP Agentic Security Initiative, and MITRE ATLAS.

Ascend AI automates the simulation of complex attack paths, attempting to force agents into misusing their integrated tools or leaking sensitive data. When an agent successfully defends against these simulated attacks, Straiker submits the assessment results through secure partner interfaces. Workday then issues the cryptographic Stamp.

“Workday has set a standards-based bar for what a trusted agent looks like, and Agent Passport gives that bar teeth,” noted James Beasley, vice president of business development at Straiker. “We are proud to be among the first partners planning to bring independent red-team results into the system of record where Workday customers make decisions about agents.”

Importantly, these Stamps are strictly tied to the specific version of the agent that underwent testing. If a developer modifies the agent's underlying model or updates its logic, the Stamp does not carry over. The application must be re-evaluated, ensuring that continuous deployment does not result in continuous vulnerability.

Venture Backing and the Politics of App Store Security

As major cloud platforms race to build the definitive "agent onboarding stack," questions regarding ecosystem governance are inevitably surfacing. Platform operators are essentially acting as the gatekeepers of enterprise AI, determining which security standards matter and who is qualified to enforce them.

The integration highlights an interesting dynamic in this emerging economy: Straiker is a Workday Ventures portfolio company. In traditional auditing environments, a financial relationship between the platform operator and the independent auditor might raise concerns about conflicts of interest.

However, the architecture of Agent Passport is explicitly designed to mitigate this risk through a strict separation of powers. The governance model dictates that the party testing an agent is never the party that vouches for it. Workday defines the specific security claims, independent partners like Straiker execute the testing against public standards, and Workday alone issues the final cryptographic Stamp.

Furthermore, the ecosystem is not intended to be a monopoly. The platform has already announced partnerships with other major players, including Cisco AI Defense for additional adversarial testing and OneTrust for governance, risk, and compliance (GRC) assessments. By building on open standards and allowing multiple independent assessors to contribute different Trust Categories—such as Build Integrity, Data Sovereignty, and AI Governance—the system provides enterprise buyers with diverse, comparable perspectives on an agent's true risk profile.

Enterprise Adoption and the Liability Landscape

For Chief Information Security Officers (CISOs) and corporate risk managers, the shift from self-attestation to cryptographic verification represents a massive operational upgrade. Historically, assessing the security of third-party enterprise extensions involved deciphering opaque bug reports and navigating lengthy legal questionnaires.

Under the new model, AI agents will effectively ship with immutable governance records. Security teams can look at a dashboard, verify the cryptographic evidence, and make immediate, informed decisions about deployment.

While industry experts caution that third-party audits and cryptographic Stamps do not provide an absolute legal shield against liability in the event of a catastrophic data breach, they do offer something arguably just as valuable: an auditable trail of rigorous due diligence. If an AI agent accesses sensitive payroll data and executes an unauthorized transaction, proving that the enterprise only deployed applications carrying verified resistance to known attack vectors will be critical in regulatory reviews and compliance audits.

The era of deploying experimental AI in the enterprise and hoping for the best is rapidly closing. As autonomous agents become deeply embedded in the digital supply chain, verifiable resilience is no longer just a technical preference; it is becoming a fundamental requirement for doing business.