Vanta to Acquire Pillar Security for AI Agent Governance

Vanta and Pillar Security announced the agreement on 9 November 2026 in two coordinated posts. Vanta’s post is titled “Vanta + Pillar: The System of Runtime Action for AI Governance”, and Pillar’s co-founders Dor Sarig and Ziv Karliner published “Vanta to Acquire Pillar Security: Building Trust in AI, Together”. Both state the same core fact: Vanta has signed a definitive agreement to acquire Pillar Security. Financial terms were not disclosed.

Vanta frames the agreement around a change in how companies use AI. In its words, “Agents have gone from answering questions to pushing code, writing customer records, and acting across business systems on their own.” Security leaders are asked to accelerate AI adoption while protecting the enterprise from the risk it carries, and Vanta says a governance program needs “the power to see and act in real time.”

A Trust Graph with over 400 integrations connects policy to enforcement

Vanta says security and GRC teams are both asked to manage AI risk but operate from different systems and different priorities. Its answer is to connect Pillar’s runtime protection and adversarial testing to the Vanta Trust Graph, which it describes as “a live map of your trust program powered by over 400 integrations.” The stated aim is to “turn governance requirements into enforceable boundaries grounded in the context of your business.”

Vanta gives a concrete example of how the combined platform will behave. If an agent attempts an unauthorised action, such as accessing restricted data or sharing sensitive financial information in an unapproved channel, the platform will block it. The Vanta Agent will then alert owners and generate an audit trail showing exactly which protections were enforced. In that description, protection and proof come from the same event, which is the idea behind the phrase “runtime action”.

Pillar’s co-founders describe the same combination from the product side. They write that the agreement brings together two things organisations need to manage as one: “how they expect AI to behave, and how they ensure it does.” Connecting Pillar’s capabilities with Vanta’s Trust Graph means, in their words, that “policies can guide testing and runtime protection, findings can reveal where controls need strengthening, and security activity can provide evidence that safeguards are being enforced.”

Pillar’s own account begins in 2023. “When we founded Pillar in 2023, most of the conversation around AI security was about chatbots and prompts,” the co-founders write. “We saw something different coming: AI would move from answering questions to taking action, with agents that access data, use tools and make decisions on a company’s behalf.” The platform they built discovers, tests and secures AI applications and agents across their lifecycle, and blocks unauthorised actions at runtime.

Pillar’s about page describes the two co-founders. Dor Sarig is a security expert with over a decade of experience building security products for Fortune 500 companies, spanning offensive security, product development and research through work with Israel’s Ministry of Defense and security companies Perimeter81 and Cymulate. Ziv Karliner has over a decade of experience in research and the development of security products, with expertise in financial cybercrime, cloud and software supply chain security, shaped by work in the Israeli intelligence forces and at IBM Trusteer and Aqua Security.

On 16 April 2025, Pillar announced $9 million in seed funding led by Shield Capital, with participation from Golden Ventures, Ground Up Ventures and strategic angels. In the same post the founders named Similarweb, Eleos Health and AvidXchange among the organisations that work with Pillar, and described one platform that provides visibility, risk assessment and risk mitigation across the full AI lifecycle, from development and testing through to real-time production monitoring, detection and response.

Pillar’s research draws on millions of real-world AI application interactions

Pillar says its research is powered by insights from analysing millions of real-world AI application interactions, developed with ecosystem partners. That programme has produced named, fixed vulnerabilities. In the seed announcement Pillar described the “Rules File Backdoor” in popular AI coding agents. On 4 February 2026 it published research on a sandbox escape in n8n, an open-source workflow automation platform that it says powers hundreds of thousands of self-hosted instances, and says the vulnerabilities allowed any authenticated user to seize complete control of the server.

On 5 May 2026 Pillar described a CVSS 10 critical vulnerability in Google’s AI-powered GitHub workflows that affected gemini-cli, a repository with 101,000+ stars. Pillar says it found the same vulnerable workflow across at least eight other Google repositories, and that Google patched the vulnerability two days after Pillar disclosed it and released a fixed version of gemini-cli. The co-founders cite this research work in the 9 November 2026 announcement as examples of critical vulnerabilities rated CVSS 10 that their team has uncovered.

SAIL passed 50,000 downloads and now maps 91 AI risks

SAIL, the Secure AI Lifecycle framework, began as a shared playbook for security, engineering and governance teams. Pillar launched SAIL 2.0 on 8 July 2026, a year after the first version, and says that more than 50,000 downloads later practitioners had shown where the framework needed to evolve next. The risk catalog grew from more than 70 risks in V1 to 91, each with a stable SAIL ID, a concrete example, the assets affected, mitigations and standards citations.

SAIL 2.0 keeps seven lifecycle phases and organises the agentic attack surface into three zones: AI assets in code and pipelines, cloud agents and endpoint agents. Its reviewers include CISOs and security leaders from JPMorganChase, Philip Morris International, Orange Cyberdefense, MyHeritage and Eleos Health, and the framework remains free and open source. It maps to the EU AI Act, ISO/IEC 42001, the OWASP Top 10 for Agentic Applications (2026), the OWASP Top 10 for LLM Applications (2025) and AIUC-1, so a team can filter the catalog by the framework it is accountable to.

Third-party recognition followed. Pillar says Gartner recognised it in the report Coolest Vendor Innovations in AI Software Security 2026, dated 1 July 2026, highlighting its RedGraph approach to offensive agentic AI security testing, its dynamic runtime guardrails and the SAIL framework. Pillar also announced a partnership with Wiz on 23 March 2026, in which Wiz discovers AI workloads in the cloud and Pillar’s RedGraph engine tests those endpoints and feeds evidence-backed results back into the Wiz Security Graph.

Vanta was founded in 2018 in the wake of several high-profile data breaches, according to its about page, with a vision to restore trust in internet businesses by enabling companies to improve and prove their security. Christina Cacioppo, Vanta’s CEO and co-founder, started the company after experiencing the manual process of getting SOC 2. Vanta is remote-first, with offices in San Francisco, New York, Sydney, Dublin and London, and lists Sequoia Capital, Craft Ventures, Y Combinator, J.P. Morgan and Goldman Sachs among its backers.

Vanta’s growth figures come from its own announcements. On 29 April 2026 the company said it took two years to grow from $10M to $100M in annual recurring revenue and 15 months to reach $200M, and that it had crossed $300M just nine months later. Vanta added that its growth rate had increased in each of the past four quarters. The workbook at the end of this report sets the $300M figure against the $200M figure so readers can see the ratio for themselves.

Customer numbers move in the same direction. In its Series D announcement of 23 July 2025 Vanta said it helps over 12,000 companies scale security and GRC, and its announcement of 9 November 2026 says over 20,000 customers rely on Vanta. In the April 2026 post the company added that 60% of the companies on the Forbes AI 50 are Vanta customers, and that the combined market cap of those customers is $560 billion.

The Series D, announced on 23 July 2025, raised $150M at a valuation of $4.15 billion. Vanta described the funding as a way to accelerate into what it called the back half of the chessboard: zero-touch security reviews, remediation through natural language and automatic mapping of contract commitments to controls. In Acquiry’s reading, the Pillar agreement extends that agenda from questionnaires and evidence to the AI agents that now act inside customer environments.

The Trust Graph links controls, vendors and evidence across 400+ integrations

Vanta describes the Trust Graph as the industry’s first real-time map of a customer’s controls, vendor relationships, evidence and compliance obligations, built on 400+ integrations and updated continuously as the environment changes. When a customer brings on a new AI tool, the Trust Graph connects that vendor’s risk profile to who has access, what data they can reach and how it maps to compliance frameworks. On top of it, the Vanta Agent acts as what the company calls a 24/7 GRC engineer, orchestrating compliance, audit, vendor risk, questionnaires and customer commitments.

Vanta shares customer results that show the model at work. According to its April 2026 post, Samsara used the Vanta Agent to consolidate 820 controls across 10 frameworks into a common controls framework of roughly 260 controls, and its vendor review process now runs 50% faster. Vanta’s Series D post adds that its AI can automatically draft more than 80% of questionnaire responses, that its AI-generated answers are accepted 95% of the time and that security reviews complete 81% faster. These are company-stated results and have not been independently verified.

Riskey, acquired in July 2025, is already part of Vanta’s vendor risk product

Vanta has integrated an acquired technology into its platform before. On 17 July 2025 it announced the acquisition of Riskey, a company it described as leading the way in real-time third-party risk monitoring. Vanta said it had already integrated Riskey’s technology into its Vendor Risk Management product, giving customers continuous monitoring of third- and fourth-party risk across vulnerabilities, breaches, misconfigurations, leaked credentials and subprocessors. Riskey co-founder Koren Molcho said the team was “incredibly excited to join this team and bring even more value to customers, together.”

Vanta announced AI Governance on 30 July 2026, built on the same foundation. In that post a Vanta customer is described as having found 900+ Copilot agents it had never inventoried, and Vanta says the Trust Graph, already running for thousands of companies, gives an inventory of agents that arrives “pre-loaded with context, not a blank row.” Early access opened with AI inventory and impact assessments aligned to frameworks such as ISO 42001 and the EU AI Act.

Vanta’s announcement names four capabilities that Pillar’s technology will bring to the platform, each with a date. The order is deliberate: the platform first sees what is running, then detects behaviour that falls outside policy, then tests how the environment responds to an attacker, and finally enforces boundaries before an unauthorised action takes place.

1. 1.AI Inventory (January 2027): automatically maps AI applications, agents, models and MCP servers and connects them to owners, approved use cases, data access, vendors, risks and controls.
2. 2.Runtime monitoring and detection (January 2027): detects suspicious activity and guardrail violations and prioritises investigations with context from the Trust Graph.
3. 3.Red teaming (Spring 2027): uncovers vulnerabilities through adversarial testing tailored to the business context, identifying exploitable paths and where protections need strengthening.
4. 4.Runtime enforcement (Summer 2027): turns governance requirements into enforceable boundaries, blocking unauthorised actions before they happen.

AI Inventory in January 2027 starts with knowing what is running

Vanta’s AI Governance post explains why visibility leads. It says most security leads cannot give a straight answer when asked how many AI agents are running inside their business, and that agents run everywhere from approved platforms to developers’ laptops to production code. An inventory that arrives already wired to the systems agents touch, the data they can reach, the vendor behind them and the controls meant to cover them lets a team rank a high-risk agent above a low-risk one on day one.

Pillar’s published red-teaming work shows what adversarial testing means in practice

Pillar launched Red Graph Suite on 6 August 2026 as a platform for the full AI red-teaming lifecycle. It maps an AI system’s attack surface, runs autonomous adversarial tests against the live application, proves what was actually executed and tracks security posture across every release. Findings carry a video of the attack, a step-by-step log of the agent’s actions and the full conversation, and the suite maps to frameworks including the OWASP Agentic AI and LLM Top 10s, MITRE ATLAS, NIST’s AI Risk Management Framework and SAIL.

Pillar says the suite scores attack success rate per objective per release and labels each result as stable, newly found, hardened or regressed. In one assessment it describes, an autonomous attacker tested an AI helpdesk agent across six versions against the same 17 attack objectives and produced a map of which defenses held, which changed and which regressed after a fix. Pillar’s point is that a point-in-time assessment describes a version of an AI system that no longer exists, because prompts, tools, knowledge bases and model updates change it weekly.

Runtime enforcement in Summer 2027 completes the loop

The last milestone is runtime enforcement, which Vanta says will turn governance requirements into enforceable boundaries and block unauthorised actions before they happen. Pillar’s Gartner post describes how its guardrails are calibrated by red-team findings: validated findings feed the guardrails at the application layer and on the endpoint, while runtime telemetry flows back into the asset map and the next round of attack simulations. Vanta’s own description of the end state is that the Vanta Agent alerts owners and produces an audit trail of the protections enforced.

Vanta states that AI Governance by Vanta will be available as a paid offering in January 2027, and its announcement says that real-time control of AI is the capability customers are asking for most. In Vanta’s words, the company is building it “starting with AI Inventory in January, and runtime enforcement behind it.”

Pillar’s Gartner post sets the commercial scene. It cites Gartner forecasting worldwide AI cybersecurity spending to nearly double in 2026, from $25.9 billion to $51.3 billion, and to reach $86 billion by 2027. The same post cites a Gartner survey in which 14% of software engineering teams have advanced skills in AI security or platform engineering, with about one-third starting to build basic proficiency. The workbook at the end of this report calculates the forecast growth directly from those inputs.

Agent adoption is the driver Pillar points to. In its Red Graph Suite announcement, Pillar cites Gartner predicting that 40% of enterprise applications will feature task-specific AI agents by the end of 2026, up from under 5% in 2025. It adds that the EU AI Act’s obligations for high-risk AI systems apply from 2 August 2026 and require evidence of resilience to attack, so an organisation needs evidence about the system it shipped last week, not only about an earlier test.

Vanta’s data shows 70% of companies already have shadow AI

Vanta’s own data points the same way. In its April 2026 post the company said 70% of companies now have shadow AI, meaning tools and models adopted without formal security review, and that LLMs are, in its words, “52% more likely to get a high risk designation in Vanta compared to traditional SaaS.” Vanta said those findings have led companies that first came to it for compliance to turn to it for third-party risk management as well.

In the 9 November 2026 announcement, Vanta says that over 20,000 customers rely on it and that “real-time control of AI is the capability they’re asking for most.” Its 30 July 2026 AI Governance post shows what that request looks like in practice: a customer discovered that an approved vendor had switched on an AI feature with access to customer data, without the team knowing.

Analyst and market language now describes governance at runtime

Analyst language has started to match the product language. Pillar’s post notes that “AI Software Security” is now a named analyst category, and IDC Research Director Phil Harris is quoted in Vanta’s announcement describing a market shifting “from AI governance on paper to governance at runtime.” Both companies use the same vocabulary of inventory, testing, monitoring and enforcement, which makes their roadmaps straightforward to compare with outside forecasts.

Everything confirmed so far is dated. The agreement was signed and announced on 9 November 2026, AI Inventory and runtime monitoring and detection follow in January 2027, red teaming in Spring 2027 and runtime enforcement in Summer 2027. Vanta says AI Governance by Vanta will be a paid offering from January 2027, which gives the first commercial milestone a fixed month.

Pillar’s co-founders frame the move as a way to reach more teams. “Joining Vanta lets us bring our work to more teams and connect it to the trust programs they already run,” they write, and they thank their team, customers, partners, investors and the security community “for helping shape Pillar and SAIL.”

Four dated milestones to track between now and Summer 2027

• –9 November 2026: definitive agreement announced by Vanta and Pillar Security.
• –January 2027: AI Inventory and runtime monitoring and detection, and AI Governance by Vanta available as a paid offering.
• –Spring 2027: red teaming.
• –Summer 2027: runtime enforcement.

For Vanta’s customers, the first capability to watch is AI Inventory, which Vanta says automatically maps AI applications, agents, models and MCP servers and connects them to owners, approved use cases, data access, vendors, risks and controls. For Pillar’s customers, the founders write that they are grateful for customers who trusted them with difficult problems, and that joining Vanta connects their work to trust programs those organisations already run.

For founders and boards of other software and security companies, the agreement is a worked example of how a specialist technology and a platform with a large customer base can combine around a single roadmap. A buyer with over 20,000 customers and a target with published research, an open framework and named partnerships can both describe, in dated terms, what changes for customers. The same discipline applies in any sector and any market: a clear statement of what the combined company will deliver, and when.