Straiker Raises $64M to Secure Enterprise AI Agents

According to a PR Newswire announcement, Straiker raised $64 million in a Series A that increases its total funding to $85 million, with the round led by Marathon Management Partners and participation from Citi Ventures, Illuminate Financial and Workday Ventures, plus continued support from Bain Capital Ventures and Lightspeed. SiliconANGLE and PR Newswire describe Straiker's product as an "agentic security" platform that performs agent discovery, pre-deployment adversarial testing and runtime protection across enterprise environments. Per PR Newswire and SiliconANGLE, Straiker's STAR Labs adversarial testing reported 36% of successful attacks on coding agents resulted in remote code execution and 91% of attacks on productivity agents led to silent data exfiltration. SiliconANGLE also includes a direct quote from CEO Ankur Shah: "Demand is outpacing anything we forecast," and reports the company said run-rate revenue has grown more than 15-fold in under a year.

Agentic workflows combine API access, credentialed system actions and multi-step planning. Industry reporting names representative integrations that Straiker supports, including Codex, Cursor, Claude Code, Microsoft Copilot and ChatGPT Enterprise (reported by ISMG/BankInfoSecurity). From a defender's viewpoint, the three-stage approach Straiker describes-discovery, pre-deployment adversarial testing, runtime enforcement-maps to three distinct technical challenges: scalable telemetry and provenance for discovery; automated adversarial generation and scoring for testing; and low-latency policy enforcement for runtime controls. Each stage has different engineering trade-offs: telemetry consistency and tagging across SaaS and cloud APIs, high-quality adversarial corpora that generalize across agent prompts, and safeguards that avoid blocking legitimate agent actions while preventing lateral movement.

Industry reporting frames agentic security as an emergent subdomain of both cloud security and ML security. PR Newswire cites IDC forecasts that more than one billion AI agents may be deployed by 2029 (as reported in the release). Independent coverage highlights recent incidents, for example, reporting that attackers manipulated an AI support agent at Meta to reset account credentials, to illustrate real-world consequences. Observed patterns in similar transitions: organizations that adopt new runtime automation platforms typically confront gaps in asset inventory, insufficient test harnesses for automated workflows and brittle enforcement when agents interact with third-party services.

Monitor technical signals and vendor coverage for:

• •standardized telemetry schemas for agent actions and provenance
• •public benchmarks for adversarial agent robustness (attack generation and detection rates)
• •emerging telemetry integrations with SIEM and cloud-native policy engines. Also watch whether customers publish post-incident disclosures or red-team reports that validate the incidence rates published by vendor labs. Finally, track how investors and competitors position around the same problem space: PR Newswire lists Marathon Management Partners and several strategic investors; enterprise uptake and case studies will clarify whether agentic security becomes a distinct procurement category or is absorbed into existing cloud/security products

Editorial analysis: For ML engineers and security teams, the near-term takeaway is practical: treating agents as first-class assets-inventory, test harnesses, and runtime policy-reduces surprise. Industry teams should evaluate how their telemetry, identity and least-privilege controls accommodate autonomous workflows and whether adversarial testing can be integrated into CI/CD or model deployment pipelines.