If you run compliance on Vanta, the AI governance module you may be testing in early access turns paid in January 2027, but blocking a misbehaving agent waits until summer 2027. Vanta signed a definitive agreement on November 9, 2026 to buy Pillar Security, an Israeli startup that tests and guards AI agents at runtime, and published a four-step roadmap for folding Pillar into its platform. Neither company disclosed the price, the closing date or what happens to Pillar's standalone product. That leaves two groups with a decision to make before January: Vanta customers choosing between Vanta's AI tier and a separate agent-security tool, and Pillar customers who have no written continuity commitment.
What Vanta Is Buying From Pillar
Vanta is buying a runtime security product and the research team behind it, and plans to run both through its compliance graph. Pillar, founded in 2023 by Dor Sarig and Ziv Karliner, discovers AI applications and agents, attacks them in testing, and blocks unauthorized actions in production. Its site lists modules for AI discovery and posture, red teaming, runtime guardrails, an "agentic endpoint" product and governance, with cloud, self-hosted and hybrid deployment options.
The research side is the part Vanta keeps pointing to. Pillar's announcement credits its team with a CVSS 10 sandbox escape in n8n and a Gemini CLI flaw that let a single GitHub issue compromise the tool's supply chain. It also says its SAIL framework for securing AI agents has passed 50,000 downloads and maps 91 AI risks across seven lifecycle phases.
Pillar is small. It raised a $9 million seed round in April 2025 led by Shield Capital and named Similarweb, Eleos Health and AvidXchange as customers at the time. Calcalist reports that the deal is worth about $150 million in cash and stock, that Pillar's 20-plus employees will become Vanta's R&D center in Israel, and that both founders will stay to run the AI security business. Neither company has confirmed the price, so treat it as a report.
The buyer is much larger. Calcalist puts Vanta at more than 20,000 customers, up from 16,000 in the first quarter of 2026, and says it passed $300 million in annual recurring revenue in April 2026, 63% growth on the year before. Those are company figures relayed by the press, not audited numbers.
When Each Piece Ships, and When You Start Paying
The roadmap arrives in four steps across 2027, and the pricing change comes before most of them. According to Vanta's announcement:
| Capability | Vanta's stated date | What it does |
|---|---|---|
| AI inventory | January 2027 | Maps AI apps, agents, models and MCP servers to owners, data access, vendors, risks and controls |
| Runtime monitoring and detection | January 2027 | Flags suspicious activity and guardrail violations, ranked using Vanta's Trust Graph |
| Red teaming | Spring 2027 | Adversarial testing tailored to your business |
| Runtime enforcement | Summer 2027 | Blocks unauthorized agent actions before they execute |
| AI Governance as a paid product | January 2027 | No price published |
Read the table from the bottom. AI Governance by Vanta becomes a paid offering in the same month that inventory and detection arrive. Red teaming comes a quarter or so later, and inline blocking comes about six months after you start paying. If you sign in January, you are buying a detection and evidence product with enforcement on a roadmap.
That matters because Vanta has been describing boundaries since the summer. Its July 30, 2026 launch post opened early access and said customers could define what each agent may do while Vanta detects activity outside those limits, with impact assessments aligned to ISO 42001 and the EU AI Act. The product page still shows a waitlist and no price. Demand is real by Vanta's own account: Pulse 2 reports that the early-access program drew about six times the demand Vanta had capacity for. Buying Pillar is how Vanta closes the distance between detecting a violation and stopping one.
What Happens to Pillar's Standalone Product?
Nobody has said, and Vanta's last deal suggests the product gets absorbed. Neither the Vanta post nor the Pillar post mentions existing Pillar contracts, support, pricing or whether Pillar will keep selling on its own. Both describe Pillar's capabilities connecting to Vanta's Trust Graph, which draws on 400-plus integrations.
The closest precedent is inside Vanta. When it bought Riskey in July 2025, the vendor-monitoring technology went into Vanta's Vendor Risk Management product, and customers were told to contact their account team for access.
The wider AI security market has followed the same pattern. When Check Point agreed to buy Lakera in September 2025, the release turned Lakera into the base of a Check Point AI security center, and the page now lists the products under Check Point names (AI Red Teaming and AI Agent Security). We have seen the same with Beacon and Haize Labs, Anaconda and Enkrypt AI and CrowdStrike and Dreadnode: a red-teaming or guardrail startup becomes a feature of a bigger platform, and the standalone buyer has to negotiate for anything else.
Two Pillar specifics deserve a direct question. Pillar sells self-hosted and hybrid deployment, and Vanta is a SaaS platform; if you run Pillar's guardrails inside your own environment, ask whether that option exists in the Vanta version. And Pillar's runtime product sits in the path of your agents' traffic. A guardrail in the request path is a dependency of your production system, so a forced migration there can take agents offline as well as reopen a contract.
The Case for Buying Agent Security From Your GRC Vendor
The strongest argument for the deal is that runtime findings and audit evidence finally land in the same system. Vanta's example is an agent that tries to reach restricted data or share financial information through an unapproved channel: the platform blocks it, alerts the owner and writes the audit trail. Vanta CEO Christina Cacioppo put it as customers, regulators and auditors wanting proof that a company's AI "does what it says on the tin," as quoted by Pulse 2. If your auditor is already in Vanta and you are working toward ISO 42001 certification, evidence that writes itself has real value.
The argument against is depth and timing. A compliance platform is optimized for proving controls exist, and a runtime guardrail is judged by its false-positive rate and the latency it adds in production. Pillar has just over 20 people, and its enforcement reaches Vanta customers in summer 2027. Competitors already ship it. Drata, another compliance platform, moved its AI Agent Governance product to limited availability on August 4, 2026, with an MCP proxy that stops violating tool calls inline. That proxy is fully supported only for agents on Anthropic today, with OpenAI, Vertex AI and Bedrock still in development. Check Point's Lakera-based AI Agent Security is sold with real-time runtime enforcement.
So you are choosing between one vendor with enforcement in mid-2027 and a separate tool that may enforce today but has to be wired into your evidence trail by hand. Either can be right for a given team. What you should avoid is paying for Vanta's tier in January on the assumption that it already blocks agent actions.
What to Do Before January 2027
Vanta customers should price the AI tier against a standalone tool now, while the early-access terms still give you leverage. Pillar customers should get continuity in writing before the deal closes.
This Week:
- If you are in Vanta's AI Governance early access, ask your account team three things in writing: the January 2027 price, whether your current access converts automatically to paid, and whether runtime enforcement is included in that price or sold later.
- If you are a Pillar customer, ask Pillar for a dated commitment on standalone availability, support terms and your renewal price through at least your next contract term, and ask whether self-hosted and hybrid deployments will continue.
- List every agent you would put under runtime control, with its owner and the data it can reach. You need that inventory whichever vendor you pick, and it tells you how many agents you would be paying to monitor.
This Month:
- Run one agent through two tools side by side: Vanta's early-access detection and one specialist runtime product. Record false positives and added latency over a week of real traffic.
- Check your change-of-control clauses. If your Pillar contract has none, ask for a termination right and data export if the product is discontinued or moved to Vanta-only pricing.
- Map who owns the decision. A GRC lead will favor the Vanta tier for evidence; a security engineering lead will favor the tool that blocks today. Get both in the room before renewal.
Before Renewal:
- Tie any Vanta AI Governance commitment to the summer 2027 enforcement date, with a credit or exit right if it slips.
- If you are buying red teaming separately, compare it against the spring 2027 Pillar capability and our guide to red-teaming agents before production before you renew a standalone contract.
The Bigger Picture
Compliance platforms are moving from documenting AI controls to running them, and security platforms are moving into AI governance from the other side. Collibra bought trail ML for governance workflows in October, CrowdStrike bought Dreadnode for testing, and Vanta is now buying runtime protection. Cloud security went through a similar consolidation, when separate posture, workload and compliance tools were bundled into platforms, and buyers of the point products had to renegotiate or migrate as each one was absorbed.
Vanta did publish dates, and they put billing in January and enforcement in the summer. Get the January price in writing, and make the summer 2027 enforcement date a contract term.
Continue Reading
- CrowdStrike Agrees to Buy Dreadnode for Automated AI Testing
- Collibra Acquires Munich Startup trail ML for AI Governance
- Beacon Bought Haize Labs. Who Red-Teams Your Agents Now?
- Upwind Acquires Aegis to Expand Into AI Agent Security
- Get ISO 42001 Certified When a Customer Requires It
- How to Red-Team Your AI Agents Before Production