What happens when the security model built for static data meets AI agents that move, copy, and act on data faster than any policy can follow? The CISOs gathered at Half Moon Bay last week had a lot to say about that.
By Shashi Bellamkonda · April 26, 2026
Cyera is not selling incremental improvements to existing data security programs. It is betting that agentic AI has made the old model obsolete, and that enterprises need a platform built to move with data rather than around it. The CISOs at Half Moon Bay are already living that problem.
Something interesting happened during a break at Cyera's Customer Advisory Board at Half Moon Bay last week. Multiple chief information security officers stopped me to thank me for raising the concept of acceptable use as a framework for thinking about AI policy. Not for a specific product feature. Not for a roadmap item. For a way of framing the problem.
That reaction told me more about where enterprise security programs actually are than any product announcement could. These are sophisticated practitioners at organizations that have invested heavily in data security infrastructure. They are not looking for more tools. They are looking for a mental model that lets them make decisions faster than the threat is moving.
Acceptable use is that model. It asks a different question than traditional security policy. Instead of "where is the data and who can see it," it asks "what are people and systems allowed to do with this data, and under what conditions." That shift matters enormously when the actors include AI agents that do not have job titles, do not sit in org charts, and do not read policy documents.
The old security model assumed data stayed put
For two decades, enterprise data security was built around location. You identified where sensitive data lived, built controls around those locations, and monitored who accessed them. That model worked reasonably well when data moved slowly and predictably, through file shares, email attachments, and scheduled database exports.
Cloud broke that model structurally. AI is now accelerating the damage. An employee pastes a contract clause into a browser-based AI tool to clean up the language. A financial model moves from SharePoint to OneDrive to a shared folder to an external summary. An AI agent retrieves customer data to complete a task, creates derived outputs, and stores them somewhere the original policy never anticipated. None of those actions go through the controls that were designed to catch them.
The security team finds out after the fact, if at all.
"Security that moves at the pace of your data. That is what the CISO room at Half Moon Bay was actually asking for."
What Ryft adds to the platform argument
Cyera's acquisition of Ryft is worth reading in this context. Ryft, founded in 2024 and backed by Index Ventures and Bessemer Venture Partners, built what it describes as a secure and automated data lake designed specifically for AI agents. The distinction matters. A data lake for AI agents is not the same as a data warehouse for human analysts. Agents query continuously, shift context mid-task, operate across chains of delegation from other agents, and leave traces that traditional audit logs were never designed to capture.
Cyera's CTO Tamar Bar-Ilan put it directly in the acquisition announcement: agents shift identity based on the task, the tools, and the chain of delegation from other agents. That is a fundamentally different identity model than anything enterprise security was built to handle. Ryft brings infrastructure designed for exactly that model.
This is Cyera's fifth acquisition in five years. The pattern is consistent: buy the capability that extends the platform's ability to follow data wherever it goes, rather than waiting to build it internally. At a $9 billion valuation with $1.7 billion in total funding, the company has the balance sheet to keep moving at that pace.
The Landing product releases signal where the real battles are
The product suite Cyera released under its new Landing format addresses five specific failure points that security teams raised repeatedly in conversations I had at Half Moon Bay. Browser Shield intercepts sensitive data entering browser-based AI tools at the point of entry, before it reaches a public model. Data Lineage for Files maps how a sensitive document spreads across systems as copies, derivatives, and summaries multiply. Classification Topics uses AI to translate technical data classifications into business language, so a security finding about "PII in shared storage" becomes "merger planning documents on a public link," a risk that leadership can actually act on.
The Model Context Protocol server is the capability that will get the most attention from enterprise architects. It lets AI tools including Claude, Microsoft Copilot, and custom-built agents query Cyera's data security intelligence using plain language. A security analyst can ask which data stores hold the most sensitive classified data, or where a specific individual's personal data appears across the environment, and get a structured, sourced answer without pivoting between dashboards. That is a meaningful operational change for teams that currently spend too much time assembling answers manually before they can begin to act.
Cyera Privacy rounds out the release by grounding privacy operations in continuous data discovery rather than static inventories. For any organization that has tried to fulfill a data subject request, the right to erasure or access under privacy regulations, using a spreadsheet that was accurate six months ago, the value proposition is immediate.
The CISOs at Half Moon Bay are not waiting for the platform to mature
What the advisory board conversations made clear is that the urgency is real. These are not early adopters exploring AI security as a future concern. They are security leaders whose environments already have AI agents running, employees already pasting sensitive data into browser tools, and data already moving in ways that existing policies never anticipated.
The maturity conversation is also real. Most organizations are still figuring out how to govern AI usage, not just block it. Blocking alone rarely holds. Employees find workarounds, adopt personal subscriptions, or use mobile devices that sit outside corporate controls entirely. The acceptable use framework resonated in that room because it offers something blocking cannot: a principled basis for decisions that scales as both the tools and the threats keep evolving.
Cyera is building toward a platform that can enforce those principles continuously. The Ryft acquisition, The Landing product releases, and the advisory board format itself are all consistent with a company that believes the window to define this category is open now and will not stay open indefinitely.
Before your next budget cycle, ask your security team to map every place an AI agent currently touches sensitive data in your environment. If they cannot answer that question within a week, you do not have a data security program designed for the world you are already operating in. Cyera is building the infrastructure to close that gap. The real question is whether your current vendor can do the same before agents become the default interface for enterprise work.
Cyera. "Cyera Acquires Ryft to Extend its Agentic AI Security Platform." Business Wire, Apr. 2026, cyera.io.
Cyera. "THE LANDING: New Product Releases From Cyera." Cyera Analyst Portal, Apr. 2026, cyera.io.
Cyera. "Cyera Raises $400M to Meet Rapidly Growing Demand for AI Security Among Enterprises." Business Wire, 8 Jan. 2026, businesswire.com.
Cyera Customer Advisory Board. Half Moon Bay, CA, Apr. 2026.
