AI COMPLIANCE COMPARE

Dispatch · Vanta, Pillar Security, AI agents, Scytale, runtime guardrails · 2026-11-10

Vanta buys Pillar Security: what runtime guardrails mean for the agent-versus-human work split

By the Signal Desk · Published · 4 min read

Short answer

On 9 November 2026 Vanta announced a definitive agreement to acquire Pillar Security, an AI security company that tests and polices AI agents at runtime. The press release dates every new capability to 2027: AI inventory and runtime monitoring in January, red-teaming in spring, runtime enforcement in summer. Nothing from the deal is on Vanta's product pages today, so the AI agent coverage score for Vanta on this site does not move. Scytale's pricing page was edited on 6 November, but its agent allowances are unchanged and the only visible difference is that a cloud assessment add-on row no longer appears in the feature table. No score moves there either.

What did Vanta announce on 9 November?

The Business Wire release says Vanta has signed a definitive agreement to acquire Pillar Security, founded in 2023 by Ziv Karliner and Dor Sarig. Pillar's product does three things Vanta's Trust Platform did not: it inventories AI agents, runs adversarial tests against them, and blocks agent actions at runtime. Vanta frames the deal as connecting the policies that say how AI should behave with guardrails that enforce what agents do. It is Vanta's third acquisition, after Riskey in July 2025 and Trustpage in January 2023, and the release says the company now serves more than 20,000 customers, up from 16,000 in the first quarter of 2026.

What is live today and what is a 2027 promise?

This is the question that matters for a buyer comparing platforms on who does the work. The release is unusually clear about timing. AI Inventory, which maps AI applications, models, agents, skills and MCP servers to owners, data access and controls, is dated January 2027. Runtime monitoring and detection is also January 2027. Red-teaming tailored to a customer's business context is dated spring 2027. Runtime enforcement, the piece that blocks unauthorized actions in real time, is dated summer 2027. Vanta says it plans to bring Pillar capabilities into its platform in early 2027.

The one number that describes something already shipping is the Vanta Agent: daily users are up 253 percent over the last three quarters, according to the release. That agent, and the AI governance offering Vanta introduced in July 2026, are the parts a customer can use this week.

Why does an acquisition not move a score here?

The scoring rule on this site is that a criterion score changes only when the vendor page behind that criterion changes. A press release about features arriving next year is not a product page. If Vanta's AI governance or agent pages change to show inventory, red-teaming or enforcement as available features, the AI agent coverage criterion will be re-read against those pages on that date, with the old and new values reported. Until then Vanta's 7.10 total stands, and so does the 0.19 gap to Scytale's 7.29.

What changed on Scytale's pricing page on 6 November?

The page's modified date moved from 2 November to 6 November 2026. The plan cards are identical: Build for under 150 employees, Scale for 150 to 1,000, Enterprise for 1,000 and up, all sold through a demo. The feature table's agent rows are identical too: Scy Agent, agentic onboarding, gap remediator and agentic policy generator on every plan; agentic policy reviewer and the Vendor Security Assessment Agent on Scale and Enterprise only; evidence readiness checks at 100, 500 and 2,000 reviews a year; agentic questionnaire responses at 30, 150 and 1,000 a year; MCP credits at 1,000, 10,000 and 50,000 a month.

What is different is an absence. The table read on 5 November listed a cloud assessment add-on on every plan. The table read today has no such row. The penetration testing add-on is still there, with one retest on Build and unlimited retests on Scale and Enterprise. The site's Scytale page and head-to-heads have been updated to drop the cloud assessment line, with the pricing page read date set to 10 November.

How should a buyer split the work after this week?

Draw the line between what an agent does unsupervised, what an agent proposes and a human approves, and what a human still does alone. On Vanta the agent side grows in 2027 if the Pillar roadmap lands; today the human still owns AI red-teaming and runtime blocking, because the product does not do them. On Scytale the agent side is defined by the allowances above and by the MCP server launched on 13 October, which is read-only. The questionnaire and evidence review caps are the practical ceiling on how much agent work a Build customer gets before a human picks up the rest.

What should you ask Vanta on a call this month?

Ask which Pillar capability reaches general availability on which date, whether existing customers pay for it, and whether the AI Inventory will cover MCP servers your own engineers run. Ask every vendor the inventory question: Scytale's table lists AI discovery and inventory on all three plans today, with continuous AI risk monitoring and the dashboard on Scale and Enterprise only.

Questions buyers ask

Does the Pillar deal change Vanta's rank on this site?

No. Vanta stays second at 7.10. The deal is a signed agreement with 2027 delivery dates, and scores here follow product pages, not announcements.

Did Scytale change its agent allowances on 6 November?

No. Evidence readiness checks, questionnaire responses and MCP credits are the same as on 2 November. The visible change is that the cloud assessment add-on row is gone from the feature table.

Sources

Keep reading

More from Dispatch