Bonfy.AI

• Microsoft could extend Purview's labeling and enforcement across the non-Microsoft applications Bonfy covers, which would remove the breadth argument for buyers standardized on that stack.
• Bonfy has published no named reference customer, so an evaluation turns on the founders' record alone until one appears.
• The accuracy claims are the vendor's own assertions, and no independent benchmark or evaluation in the cited record could confirm them or fail to reproduce them.
• No completed attestation appears on the surfaces probed on 30 August 2026, and regulated buyers who require a readable report before signing may stall procurement.
• Contextual Data Enforcement works by replacing a native AI connector, so a platform vendor that restricts or redesigns those connector paths would cut off the deployment route Bonfy describes as a simple swap.

Bonfy.AI sells to security and compliance leaders whose organizations adopt AI faster than their data controls can follow. The company's own framing is that AI systems and agents move information through multi-step automated workflows legacy data-loss and posture tools were never designed to see or control, and that fears around data leakage intensify as organizations adopt technologies such as Microsoft 365 Copilot.

The quantified evidence for that pain stays at category level. Bonfy's sponsored announcement of the March 2026 platform release, published on Help Net Security, carries Gartner projections that 22% of cyberattacks and data leaks will involve generative AI by 2028, and that through 2029 over half of successful attacks against AI agents will exploit access-control issues. Those are market-wide forecasts the vendor relays, so the record still carries no measurement of exposure at buyers in the sectors it targets, namely financial services, healthcare, insurance, and legal. [s1, s2, s11]

Bonfy Adaptive Content Security inspects the content itself, which is the line the company draws against products that focus on model jailbreaks or agent configuration. Entity-aware engines and Adaptive Knowledge Graphs score risk in human and AI-generated content, then block, redact, or flag it across Microsoft 365, Google Workspace, Salesforce, Slack, on-premises file stores, and AWS S3.

The March 2026 release moved that engine into the agent path. Bonfy runs an MCP server an agent calls during its reasoning to evaluate content against policy and decide whether to revise, redact, block, or route for review. Contextual Data Enforcement replaces a native AI connector so each retrieval request is inspected before the model receives it, and a browser extension covers sensitive data flowing to AI tools the organization has not sanctioned.

External validation is what the record lacks. Bonfy documents its architecture on its own pages, and no third-party benchmark, open-source component, or independent technical evaluation in the cited record tests the accuracy the product claims. A buyer weighing the false-positive claim has only Bonfy's own description to weigh it against. [s2, s7, s8, s1]

Bonfy now states its relationship with Microsoft as a division of labour rather than a contest, saying Purview governs Microsoft while Bonfy governs the whole enterprise. That framing also names the vendor that could extend the same labeling and enforcement outward across other applications.

Bonfy's answer is breadth. One policy and automation engine spans email, SaaS apps, browsers, on-premises file stores, and AI agents, and the knowledge graphs add each customer's own business context on top. Bonfy supports that breadth claim with its own account of what it covers, and the cited record carries no independent comparison against other AI data-security products. [s1, s11, s2]

Bonfy left stealth in June 2025 and has shipped steadily since, reaching a second platform version on 19 March 2026 that added Google Workspace parity, agent-level enforcement through an MCP server, a shadow-AI browser extension, and completed SOC 2 Type 2 certification. Its awards page names five industry recognitions, among them Cyber Defense Magazine's Top InfoSec Innovator award for 2026 in AI data leakage prevention and a 2026 Cybersecurity Stars award for AI security.

What the record still does not carry is a customer. The cited pages name none and describe only a case study about an unnamed global financial institution that used the product on its Microsoft Copilot instance, and the launch release quotes the CISO of Cast and Crew endorsing the approach without stating that the company runs Bonfy.

Two indirect signals point at traction Bonfy has not published. TLV Partners led the seed with Saban participating, and the founders bring enterprise sales records from Skybox Security and Idaptive. Those raise the odds of early deployments, and the motion stays unproven in public while no customer is named. [s6, s12, s7, s13, s4]

Gidi Cohen founded and ran Skybox Security, a cyber-analytics firm whose technology a CTech article says was deployed across over 500 Global 2000 companies before private equity acquired it in 2018 for $375 million. Danny Kibel was CEO of the identity firm Idaptive, acquired by CyberArk, where he then led global research and development.

Skybox itself shut down in 2025, when Tufin acquired its business and technology assets, seven years after that acquisition. The 2018 sale is the exit the record carries for Cohen.

The public leadership page lists the two founders and no other executives, so the depth of the wider executive, engineering, and go-to-market bench is not visible in the cited record. [s4, s10, s3]

Bonfy states that it completed SOC 2 Type 2 certification alongside its March 2026 platform release, in its own announcement and in the sponsored version of that announcement published on Help Net Security. For a company selling into financial services, healthcare, and insurance, that is the strongest compliance signal the record carries, and it rests on the company's own word.

A probe on 30 August 2026 found the trust and security subdomains failing DNS alongside a control subdomain that also failed, the /security, /trust and /compliance paths returning 404, and no attestation badge file in the homepage markup. The product page still describes the controls as SOC 2-aligned and offers its certifications and standards on request, so a buyer confirms the certification through the vendor rather than from a published artifact.

Around that, the platform carries role-based access control, multi-factor authentication, audit logs, an optional bring-your-own-cloud deployment, and data minimization through short retention windows and content obfuscation. [s2, s9, s11, s13]

Add analyzed competitors to compare them side by side with Bonfy.AI.

Bonfy targets regulated, security-conscious buyers where AI adoption collides with compliance risk. Its own pages name financial services, healthcare, insurance, and technology, and its sponsored March 2026 release announcement names CISOs, CIOs, security architects, and AI platform teams in financial services, insurance, technology, biotech and pharmaceutical, and healthcare companies as the people who run it. An independent FinTech Global report of the seed round records the same focus on healthcare, finance, and legal.

The segmentation matches the product's controls and stays aspirational in public. The cited pages name no buyer in any of those sectors, and its single customer reference is a case study about an unnamed global financial institution.

Bonfy's stated differentiator is that it inspects content rather than models or agent configuration, applying entity-aware detection and Adaptive Knowledge Graphs that learn each organization's own structure, customers, and policies from its business applications. The engine scores risk in human and AI-generated content and can block, redact, or flag it across email, SaaS apps, file stores, and AI agents.

The March 2026 release put that engine inside the agent's reasoning loop. Bonfy runs its own MCP server that an agent calls mid-task to evaluate content against policy and then decides whether to revise, redact, block, or route the content for review.

The claimed advantage over legacy data-loss tools is contextual precision, understanding who data belongs to rather than which keywords appear. No independent benchmark, open evaluation, or customer technical writeup appears in the cited record, so the accuracy edge is asserted rather than demonstrated.

Bonfy sells top-down to enterprise security and compliance buyers through a demo-and-contact motion, with no self-serve or trial surface on the cited pages. The motion leans on founder credibility, a steady release cadence, and conference visibility, and its awards page names five industry recognitions.

Proof of demand is the gap. The cited pages name no customer and describe only a case study about an unnamed global financial institution, so the effectiveness of the motion is inferred from reputable backing and founder pedigree rather than shown.

No pricing page, published rate, or self-serve tier appears on the cited vendor pages, and the buying path they present is a demo request.

What Bonfy charges by, whether seats, data volume, or connected systems, is not publicly legible, so a buyer cannot size the product before contacting sales.

Bonfy delivers as SaaS in multi-tenant and single-tenant forms, with an optional bring-your-own-cloud deployment. Operational controls include role-based access control, multi-factor authentication, audit logs, and data-minimization options such as short retention windows and content obfuscation.

Contextual Data Enforcement deploys by replacing a native AI connector rather than by adding a gateway. Bonfy supplies the platform and the policy engine, and the customer's team configures policy and operates it.

Bonfy states that it completed SOC 2 Type 2 certification alongside its March 2026 platform release, in its own announcement and in the sponsored version of that announcement published on Help Net Security. For a company selling into financial services, healthcare, and insurance, that is the strongest compliance signal the record carries, and it rests on the company's own word.

A probe on 30 August 2026 found the trust and security subdomains failing DNS alongside a control subdomain that also failed, the /security, /trust and /compliance paths returning 404, and no attestation badge file in the homepage markup. The product page still describes the controls as SOC 2-aligned and offers its certifications and standards on request, so a buyer confirms the certification through the vendor rather than from a published artifact.

Bonfy's reach spans Microsoft 365 including Mail, SharePoint, Entra, Purview and Copilot, Google Workspace across Gmail, Drive and Directory, Salesforce, HubSpot, Slack, on-premises and cloud file stores including SharePoint and S3, and custom applications through an SDK and APIs. For AI, it ships an MCP server and agent-framework support so agents can call it during reasoning.

The March 2026 release brought Google Workspace to parity with the Microsoft integrations, so one policy plane now spans both ecosystems as well as agent workflows. That breadth is the structural argument Bonfy makes against a single-suite control, and it also deepens the integration surface a buyer would unwind on the way out, which the cited record does not size.

Bonfy's team is led by two founders with in-domain exits. Gidi Cohen founded Skybox Security, whose technology a CTech article says was deployed across over 500 Global 2000 companies before private equity acquired it in 2018 for $375 million. Danny Kibel ran Idaptive to a CyberArk acquisition and then led global research and development at the acquirer.

Skybox itself shut down in 2025, when Tufin acquired its business and technology assets, seven years after that acquisition. The 2018 sale is the exit the record carries for Cohen.

The public leadership page lists the two founders and no other executives, so the depth of the executive, engineering, and go-to-market organization is not visible in the cited record.