Prompt Security

• SentinelOne could retire the standalone Prompt Security brand and fold its screening into Singularity bundles, ending the independent roadmap. The about page is already SentinelOne branded, positions Prompt Security as built into the Singularity Platform, and routes demo requests to a SentinelOne landing page.
• Buyers running CrowdStrike or Microsoft endpoints could drop the product now that it belongs to a rival endpoint platform, shrinking the vendor-neutral market the browser and proxy deployment served.
• Palo Alto Networks, Tenable, and Cisco bought comparable AI-security capabilities, so suite pricing from those platforms could undercut any separately priced Prompt Security offering.
• Gartner Peer Insights carries critical reviews citing false positives, false negatives, and weak context awareness, so unaddressed efficacy gaps could erode the runtime-screening value as buyers test it at scale.
• The MCP risk-scoring claim over more than 13,000 public GitHub servers rests on replicable data, so registry or gateway vendors could standardize MCP security before SentinelOne invests further in that module.

Prompt Security sells to enterprise security teams that want employees, developers, and AI agents to use generative AI without leaking data or absorbing prompt-based attacks. The employees solution page leads with discovery and monitoring of every AI tool in use to eliminate shadow AI and surface the riskiest apps and users.

Independent reporting corroborates the pain and the budget behind it. CTech describes hackers crafting prompt-based exploits while employees expose sensitive data through tools like ChatGPT, the 2025 wave of nine-figure acquisitions priced the budget line, and Series A coverage placed the company protecting applications, employees, and customers across enterprises rather than early adopters. [s2, s8, s9, s12]

The product inspects prompts and responses inline and enforces policy at the point of AI use. Solution pages describe four surfaces. Employee coverage pairs shadow-AI discovery with redaction, homegrown applications get protection from prompt injection, jailbreaks, denial of wallet, and remote code execution, agentic systems get machine-level inspection through a lightweight agent or reverse proxy, and automated red teaming feeds findings into runtime protection.

The most specific capability claim is MCP risk scoring that dynamically assesses over 13,000 MCP servers on GitHub. Independent user feedback now tempers the marketing. Gartner Peer Insights reviewers post a favorable average, but the platform also draws criticism for false positives, false negatives, and detections that are not context aware, and no public documentation portal appears in the reviewed pages, so a buyer cannot judge detection quality from public pages alone. [s2, s3, s4, s5, s14]

Prompt Security competed in runtime GenAI security, the category platform vendors consolidated heavily through acquisitions in 2025. CTech reported Palo Alto Networks paying $700 million for Protect AI, Tenable paying $100 million for Apex, and Cisco buying Robust Intelligence for $500 million a year earlier, before SentinelOne reached for Prompt at a reported $250 million, and it named Lasso, Aim, and Pillar as the independents still fielding offers within days.

Inside SentinelOne the product competes platform against platform. Rival platforms made the same buy-side move, so suite bundling from those platforms is a growing competitive pressure for any buyer weighing a standalone control. The browser-extension and proxy deployment that let Prompt deploy on any stack is the asset platform bundling now puts in question. [s8, s9, s10]

Named customers speak publicly for the product. The customer page carries attributed testimonials from 10x Banking's CISO and Cymulate's CTO and publishes case studies, the homepage adds HiBob's global CISO, and the founder's acquisition announcement on the company blog reports protecting millions of AI interactions monthly inside a multi-million-dollar ARR business, a company-reported figure.

Scale stayed modest relative to the price. CTech records about 50 employees, mostly developers, two years after founding, and Gartner Peer Insights shows roughly twenty enterprise reviews across several industries. The deal price doubles as a traction signal, since SentinelOne paid a nine-figure sum for a company that had raised $23 million. [s1, s7, s9, s10, s12, s14]

Itamar Golan and Lior Drihem founded the company in August 2023 after building AI capabilities at Check Point and Orca Security, where they worked together. CTech describes both as graduates of Israel's Unit 8200, and the homepage describes the team as core members of an OWASP research team.

The two-year path to a nine-figure exit is the team's verifiable execution record. CTech reported the deal at a valuation near $250 million, and the founders' background sits at Check Point and Orca rather than a prior independent company, so credibility comes from the built product and the price an acquirer paid. [s1, s8, s12, s13]

Public trust collateral is thin. No SOC 2 or ISO attestation appears on the public pages, and the trust subdomain served a Cloudflare error page rather than a readable portal when probed in July 2026.

Named customers in regulated industries substitute in part. A CISO at 10x Banking describes production use, the customer wall shows a named bank in the Ion Bank logo, and Gartner Peer Insights reviewers span energy and utilities and retail. After the September 2025 close, the product is sold and supported under SentinelOne, so diligence increasingly runs through the acquirer rather than the standalone startup. [s7, s13, s14, s15]

Add analyzed competitors to compare them side by side with Prompt Security.

Prompt Security sells to the enterprise security team that wants employees, developers, and AI agents to use generative AI without leaking data or absorbing prompt-based attacks. The homepage frames the target as securing AI everywhere it matters, and the employees page makes the entry promise concrete: detect and monitor every AI tool in use to eliminate shadow AI and surface the riskiest apps and users.

The evidenced segments skew to regulated and large buyers. The customer page carries an attributed testimonial from a 10x Banking CISO and shows a named bank in the Ion Bank logo, the homepage adds HiBob's global CISO, and Gartner Peer Insights reviewers span energy and utilities and retail, the population where AI output and data exposure carry direct liability.

The segmentation spans four surfaces rather than one buyer. Solution pages address employee AI use, homegrown applications, agentic systems, and red teaming, so the company addresses any organization adopting AI regardless of which models it runs, which widens reach but places it against both neutral rivals and the endpoint and network platforms now shipping their own controls.

The product inspects prompts and responses inline and enforces policy at the point of AI use. Solution pages secure homegrown applications from prompt injection, jailbreaks, denial of wallet, and remote code execution, pair employee coverage with shadow-AI discovery and redaction, and feed automated red-teaming findings into runtime protection.

The agentic coverage is the most specific capability claim. The agentic page describes endpoint-level enforcement through a lightweight agent or reverse proxy and MCP risk scoring that dynamically assesses over 13,000 MCP servers on GitHub, paired with deep inspection of user-to-server interactions.

Independent feedback now tempers the marketing. Gartner Peer Insights reviewers post a favorable average, while others report false positives, false negatives, and detections that are not context aware, and no public documentation portal appears in the reviewed pages, so a buyer cannot judge detection quality, architecture, or integration depth from public pages alone.

Go-to-market runs through named enterprise references and a category that press and acquirers place without vendor coaching. The customer page carries an attributed testimonial from a 10x Banking CISO, the kind of named proof that opens regulated-enterprise conversations rather than design partners alone, and the founder's acquisition announcement on the company blog reports a multi-million-dollar ARR business, company-reported, protecting millions of AI interactions monthly.

The strongest go-to-market signal is the exit itself. CTech first reported the deal at a valuation near $250 million, while SentinelOne's later filing recorded the closing consideration at roughly $133.6 million in cash plus stock and options, and the acquisition itself doubles as demand validation for a young company sitting inside a wave of comparable buys that priced the whole category.

The motion now routes through the acquirer. The about page presents the product as Prompt Security from SentinelOne and sends demo requests to a SentinelOne landing page, so the independent pipeline that sold to buyers regardless of endpoint vendor now flows through one platform, the open question for whether its vendor-neutral reach holds.

Prompt Security publishes no public price. No Prompt Security rate card appears in the reviewed pages and the demo page is the entry point, which signals a negotiated enterprise motion rather than self-serve tiers, consistent with selling into banks and large enterprises through direct conversations.

The metering unit is not publicly answerable from the fetched record. Because the product inspects every prompt and model response inline, cost would plausibly track the volume of AI traffic screened, the same consumption logic the cloud guardrail services use, but that is inference rather than a published unit.

The absence of a public price suits the named buyer and raises a barrier for smaller teams. A bank or large enterprise folds the cost into a negotiated agreement, while a developer evaluating the product has no transparent entry point, and after the acquisition the commercial terms fold into the acquirer's selling motion rather than a standalone rate card.

Prompt Security is delivered as an inline control the customer routes its AI traffic through, configuring policy for the inline inspection the product performs. The product inspects every prompt and model response, and the agentic page offers a lightweight agent or a reverse proxy for endpoint-level enforcement, so a security team chooses the deployment that fits its environment.

The deployment flexibility is the operational differentiator. Offering an agent and a proxy lets the product sit in front of employee AI use, homegrown applications, and agent traffic without depending on a single endpoint platform, the architecture that let it deploy on any endpoint stack.

Operations carry real tuning cost. The product pairs shadow-AI discovery with redaction, shields homegrown apps from injection and denial of wallet, and scores MCP servers to flag emerging agent threats, but Gartner Peer Insights reviewers describe detection rules that are hard to tune and false-positive heavy, so the operational promise depends on staffing that tuning.

Public trust collateral is thin. No SOC 2 or ISO attestation appears on the public pages, and the trust subdomain does not resolve to a readable portal as of June 2026, so a security reviewer starts from sales conversations rather than downloadable artifacts.

Named customers in regulated industries substitute in part. A 10x Banking CISO gives an attributed testimonial describing production use, the customer wall shows the Ion Bank logo, and Gartner Peer Insights reviewers include energy and utilities and retail enterprises, external signals a buyer can weigh against the absent attestations.

After the September 2025 close, the trust question moves to the parent. The about page presents the product under the acquirer's name and routes demo requests to its landing page, so the product is sold and supported under SentinelOne and diligence increasingly runs through the acquirer rather than the standalone startup, a structural fact about where evaluation now happens rather than a capability the product owns.

Prompt Security is built to be the runtime control point an enterprise places over the AI it adopts, and the four-surface design is the strategy. It sits between users and tools, between homegrown applications and their models, and between agents and the MCP servers they call, so a security team governs employee use, applications, and agents from one place.

Outward, the company extends its control point through deployment flexibility rather than a third-party platform. The lightweight agent and reverse proxy place its enforcement inside environments the buyer already runs, which let it deploy across endpoint stacks as a neutral layer rather than a feature of one stack.

Inward, that neutrality is exactly what the acquisition puts in tension. The product now belongs to one endpoint platform, so the strategic value of being model-agnostic and stack-agnostic depends on whether buyers who run rival endpoints keep it on their shortlist, a position the acquirer's ownership names as a structural factor rather than a capability the product can defend on its own.

Prompt Security was built by Itamar Golan and Lior Drihem, Unit 8200 graduates who developed AI capabilities at Check Point and Orca Security before founding the company in August 2023, and sold it to SentinelOne with the deal completing in September 2025.

The execution record is the team's clearest signal. The founders shipped a working product spanning employee AI use, homegrown applications, agents, and red teaming, and reached an exit CTech valued at a reported $250 million, which Globes framed as an impressive cybersecurity exit reached without large Israeli venture funds, evidence of build-and-sell capability rather than a stated pedigree.

What the record adds beyond the exit is two veterans with notable prior roles at Check Point and Orca Security. Coverage centers on the product, that founder background, and the acquisition, so credibility comes from the built product and the price an acquirer paid alongside that pedigree.