An AI tool that prepares a document is useful. An AI agent that changes customer records, sends information or acts across business systems raises a different management question: who authorised the action, and what would stop it going wrong?
On 9 November, Vanta announced a definitive agreement to acquire Pillar Security. Pillar separately confirmed the agreement the same day. This is an announced transaction; those statements do not establish that completion has occurred. [1][2]
Calcalist’s CTech reported an approximate US$150 million consideration in cash and shares. That is the publication’s reported figure, rather than a price disclosed in the two company announcements reviewed for this article. [3]
The commercial significance extends beyond the transaction itself. Our reading is that governance software suppliers see an opportunity to connect a business’s AI rules with controls operating while software takes action.
Check the timetable behind the promise
Vanta’s published roadmap targets AI inventory and monitoring for January 2027, adversarial testing in spring and runtime enforcement in summer. These are planned releases, rather than evidence that the combined offering is fully available today. [1]
For a prospective customer, that distinction matters. A demonstration should establish which protections work now, what integration is required and which capabilities depend on future delivery. Commercial commitments should reflect the functionality being purchased.
The broader issue already features in UK official advice. In its interim guidance dated 20 August 2026, the National Cyber Security Centre recommends proportionate safeguards, technical restrictions, human oversight and monitoring. It explicitly advises against relying on prompting alone. This is security guidance, rather than a new legal requirement created by the acquisition. [4]
A practical starting point for founders
Consider an illustrative SME using an agent to help prepare customer proposals. Initially, the agent might read approved material and produce a draft. Giving it permission to update prices or email customers creates additional consequences. The business should make a deliberate decision before extending that authority.
Start with a simple register: which agents are in use, who owns each one, which systems it can reach and which actions require human approval. Ask suppliers to demonstrate how those boundaries are enforced, how an unexpected action is detected and who can suspend access.
Then test a realistic failure. If an agent receives misleading instructions in a document, can it expose confidential material or make an unauthorised change? A controlled exercise can reveal more than a confident presentation. Retain the results and record any remediation.
For management teams preparing for funding or a sale, our assessment is that this evidence could strengthen technology due diligence. It would give a prospective capital partner something concrete to examine alongside the productivity case. That is a practical inference, not a claim that lenders universally require an AI assurance pack or will offer better terms.
Sources & further reading
- Vanta: Vanta + Pillar: The System of Runtime Action for AI Governance ↗Published 9 November 2026
- Pillar Security: Vanta to Acquire Pillar Security: Building Trust in AI, Together ↗Published 9 November 2026
- Calcalist / CTech: Vanta to acquire Israeli AI security startup Pillar for $150 million ↗Published 9 November 2026
- UK National Cyber Security Centre: Managing the cyber risk of agentic AI ↗Published 20 August 2026
Let’s talk
Chris McDonald’s perspective
I would encourage founders to approach AI adoption with the same commercial discipline they bring to any important delegation. Be clear about the task, the authority being granted and the person who remains accountable. A useful board discussion would examine both the benefit and the controls: what time is being saved, what information is exposed, and what happens when the system makes a mistake? That conversation should involve the people responsible for operations and customer relationships as well as technology. The ambition should be to build confidence through evidence. A business that can explain how its AI works, where its authority ends and how problems are handled has a firmer basis for deciding when to expand its use.