AI Guardrail Vendors After the Acquisition Wave: What Happens to Lakera (Check Point), Prompt Security (SentinelOne), Aim Security (Cato), and CalypsoAI (F5)

Every cell in the bottom two rows reads the same way for a reason: none of these four acquirers has published a standalone rate card or a fully dated, feature-complete integration timeline for the acquired AI guardrail capability. Treat the rest of this comparison as a guide to what to ask your account team, not a substitute for asking it.

Lakera built an AI-native security platform aimed specifically at agentic AI applications, combining real-time runtime protection against prompt injection and unsafe agent behavior with continuous automated red teaming. Lakera's Gandalf red-teaming game, which crowdsourced adversarial prompt attempts against its own models, was one of the more widely known pieces of public AI security research to come out of a commercial vendor. Check Point's stated rationale was combining Lakera's runtime protection with its own Infinity architecture to secure what it calls the full lifecycle of AI: models, agents, and data.

Prompt Security focused on generative and agentic AI security for the enterprise: real-time visibility into which AI tools employees and systems were actually accessing, what data was being shared with them, and automated enforcement to stop prompt injection and data leakage before it left the organization. SentinelOne's stated goal is extending its existing Singularity Platform, built around AI-native threat detection for endpoints and cloud workloads, to cover this same generative and agentic AI surface.

Aim Security, founded in 2022 and backed by YL Ventures and Canaan Partners, addressed three related but distinct problems: securing employee use of public AI applications (the shadow-AI problem), protecting AI systems an organization builds and runs itself, and managing security across the AI development lifecycle rather than only at runtime. Cato's stated rationale was extending its SASE platform, which already sits inline on enterprise network traffic, to cover AI-specific risk using that same vantage point.

CalypsoAI, founded in 2018 with offices in Ireland and New York and backed by Paladin Capital Group, Lockheed Martin Ventures, and Hakluyt Capital, built two connected capabilities: agentic red-teaming at scale, meaning automated adversarial testing against models, applications, and agents before deployment, and runtime guardrails that detect prompt injection and jailbreak attempts in production, an approach CalypsoAI described as creating an "inference perimeter" around a deployed model. F5's stated goal is folding both pieces into its existing Application Delivery and Security Platform, the same platform that already handles application and API traffic for many of F5's large enterprise customers.

Of the four, Cato Networks has been the most specific in public statements about timing, and what it said should set expectations for the other three as well. Cato stated that Aim Security's capabilities would be offered as part of its SASE Cloud platform beginning in early 2026, and that existing Aim customers would get a migration path from the standalone product to the integrated platform capability. Read plainly, that is an admission that the standalone Aim product does not simply continue running unchanged forever: customers are being moved, on a timeline the acquirer controls, into the parent platform.

SentinelOne closed its Prompt Security acquisition on September 5, 2025 and has described the integration as extending the Singularity Platform's existing AI-native architecture to cover generative and agentic AI use. Check Point has described integrating Lakera's runtime protection into its Infinity architecture, and reporting from industry roundups (not Check Point's own press materials) describes Lakera's product and its Gandalf red-teaming community being folded into offerings branded Check Point AI Red Teaming and AI Agent Security. F5 completed its CalypsoAI acquisition on September 26, 2025 and introduced two explicitly branded products, F5 AI Guardrails and F5 AI Red Team, folding CalypsoAI's runtime detection and red-teaming capability into F5's Application Delivery and Security Platform.

What none of the four acquirers has published is a fully dated, line-item feature parity list comparing the standalone product as it existed before the acquisition to what actually ships inside the parent platform today. For a buyer evaluating any of these right now, the deployment question is less "how do I install this" and more "exactly which of the standalone product's features are live inside the parent platform as of today, versus still on a roadmap." Get that answer in writing, with dates, rather than accepting a general statement that integration is "underway."

The most significant operational change across all four deals is that the AI guardrail capability is no longer something you can adopt in isolation. Adopting Check Point's AI security capability now means being a Check Point Infinity customer, or becoming one. Adopting Prompt Security's capability means being on SentinelOne's Singularity Platform. Adopting Aim Security's capability means being on Cato's SASE Cloud. Adopting CalypsoAI's capability means being an F5 Application Delivery and Security Platform customer. For a security team already standardized on one of these four vendors for endpoint detection, SASE, or application delivery, this is a genuine operational win: one console, one contract, one vendor relationship covering both the existing platform and the AI guardrail layer, instead of a fifth or sixth standalone security tool to integrate and staff for.

For a team that is not already a customer of the relevant acquirer, the calculation is very different. Buying CalypsoAI's former capability today effectively means evaluating and buying into F5's ADSP, not just an AI guardrail. Buying Aim Security's former capability means evaluating and adopting Cato's SASE architecture. That is a materially bigger and more expensive decision than the one a buyer was making when Lakera, Prompt Security, Aim Security, and CalypsoAI were pitched as narrowly scoped, fast-to-deploy point products, which was a meaningful part of their original appeal to security teams trying to get ahead of AI risk without a platform migration project attached. Operationally, budget more evaluation time for the parent platform itself, not just a demo of the AI-specific feature, since that platform is now a mandatory part of the purchase.

None of the four acquisitions comes with published, comparable pricing, and that was already true of the standalone products before the acquisitions. SentinelOne's own disclosure gives the clearest number of the group: the Prompt Security deal was valued at roughly $250 million, made up of approximately $133.6 million in cash, 1,555,099 shares of Class A common stock, and 415,109 assumed options, as reported in SentinelOne's SEC filings. F5 has disclosed a $180 million price for CalypsoAI, with a final closing cash payment of $145.2 million. Check Point has not published an official deal value for Lakera; press estimates put the figure at roughly $300 million, and that number should be treated as a reported estimate, not a confirmed figure from Check Point itself. Cato Networks has not disclosed financial terms for the Aim Security acquisition at all.

Those are the acquisition prices, not what a customer pays for the resulting product, and none of the four companies has published customer-facing pricing for the integrated AI guardrail capability either. All four now sit inside a broader platform's enterprise sales motion: Check Point Infinity, SentinelOne Singularity, Cato SASE Cloud, and F5's Application Delivery and Security Platform. Expect a custom quote scoped to your existing or prospective contract with that vendor, and expect the AI guardrail capability to be positioned as an add-on module, a bundled feature at a higher license tier, or a distinct SKU, depending on the vendor and your negotiating leverage. Ask directly which of those three it is before assuming a number.

Each of the four acquisitions has a plausible strength case and an equally real limit, and the honest answer for all four is that it is too early to fully validate either from the outside.

Check Point's strength case is combining Lakera's runtime protection and red-teaming pedigree, including a team with backgrounds at Google and Meta, with an established enterprise security distribution channel and Check Point's existing Infinity architecture. The limit is that Check Point has not published a specific, dated feature-parity commitment, and the reported $300 million price tag is unconfirmed by Check Point itself.

SentinelOne's strength case is folding Prompt Security's real-time AI-tool visibility and enforcement into an already AI-native detection platform with a large existing endpoint and cloud customer base, giving the combined product a single-pane-of-glass advantage for existing SentinelOne shops. The limit is the same as the others: no published, itemized list of what shipped immediately at close versus what is still being engineered in.

Cato's strength case is the most concretely dated of the four: Aim Security's three-pronged coverage (public AI app use, internal AI systems, AI dev lifecycle) folded into a SASE platform that already inspects enterprise network traffic inline, with a stated early-2026 general availability date. The limit is that same date; a buyer evaluating this today, in September 2026, needs to confirm the integration actually reached that milestone and is generally available now, not still rolling out.

F5's strength case is pairing CalypsoAI's agentic red-teaming and inference-perimeter runtime detection with a platform that already handles application and API delivery for many large enterprises, under clearly branded new products (F5 AI Guardrails, F5 AI Red Team) rather than a vague integration promise. The limit is that CalypsoAI was a comparatively young, VC-backed company (founded 2018, having raised roughly $23 million as of 2023) being absorbed into a much larger public company's roadmap, and prioritization decisions inside F5's broader product portfolio are now out of the acquired team's direct control.

None of these four is the right choice for a team that specifically wants a narrowly scoped, fast-deploying, single-purpose AI guardrail product with no dependency on a much larger platform contract, because that product category, as these four vendors originally defined it, effectively no longer exists among these particular names. A team in that position should look at whichever independent AI guardrail vendors have not been acquired as of this writing (HiddenLayer, WitnessAI, Pillar, and Noma among the names we have covered in prior comparisons) rather than assume one of these four will still behave like a standalone point product a year from now.

The right way to evaluate these four today is almost entirely a function of what platform you are already committed to, or willing to commit to, rather than a feature-by-feature comparison of the original standalone products.

Because the acquisition changes what you are actually buying, the standard AI guardrail proof-of-concept checklist needs three additional items specific to a post-acquisition product.