Defiance AI Security™
Secure Every AI Interaction
Read the WHY behind every prompt, response, and agent action to give the evidence you need to govern AI
Interactive AI Security | Holistic Control
Visibility
Gain visibility to AI use and independent agentic behaviors. Understand the ‘what’, ‘who’ and ‘where’ of token use to uncover unseen risks.
Agentic Identity
See what agents are up to! Connecting agentic activity to how user prompts are enacted downstream into system actions.
Secure AI Usage
Proactively listen to AI and receive distillations on usage – both good and bad. Verify how agentic usage is happening to evolve policies and guide business efficiencies.
Risk Mapping
Understand the real risks the agentic plane introduces to your organization. See “under the hood” to how malicious AI activity enters and infiltrates.
Policy & Governance
Turn governance policy into clear enforceable controls. Govern which AI tools are used and how they are permitted to run, throughout the agentic landscape.
Expert Support
A collaborative and living approach to development of AI policy – means your organization gets ongoing support and real time informed governance that keeps up with the blazing pace of AI change.
Threat Mitigation
Detect and stop threats in real time. Inspect what happened, map full impact/reach and deploy containment actions on malicious activity before it spreads
Operationalize AI Security
Secure the entire agentic plane – across all AI interactions throughout your organization. Ensures secure operations for models, agents, users, endpoints, identities, assets with technology that only gets smarter and keeps your policy enforced.
Rogue AI Is On The Rise
“This is not classic IT observability into whether something happened but rather contextual visibility into why something happened.”
Full Visibility | Interactive AI Security
Traditional cybersecurity and AI firewalls see the transaction. Interactive AI security adds the context – understanding intent, behavior over time, agent activity, and potential consequences.
Easily Enforceable AI Governance
Run your own or simply use the AI governance and policies that come with the Defiance platform. Gain transparency to policies, what was attempted, and points of enforcement to evolve your AI policy as your business matures in AI adoption.
Deploy Where Your Data And Risk Live
AI Security Edge (AISE)
Endpoint EXE or SASE plug-in
Teams deploying standardized models (Claude, OpenAI, Gemini) for chat and co-work, or adding agentic workflows. MDM-deployable, no SASE prerequisite.
What's Included
- Endpoint EXE or SASE plug-in, proxying to your shared edge appliance.
- AI security, attack-surface, and tool-surface posture (SPM · ASM · TSM).
- Inline blocking as an optional add-on
Enterprise
Gateway on your network stack
Mature security environments processing sensitive data with AI, building or fine-tuning their own models, or running customer-facing AI in regulated industries.
What's Included
- Everything in AISE, with inline blocking standard.
- Attaches to your SASE, firewall, colo, or cloud VPC.
- BYOK key custody, and it works with your existing DLP tools.
NonPublic
Fully air-gapped, one-way CTI
Air-gapped environments where cloud LLMs are legally off-limits. This includes IC, DoD, CMMC, FedRAMP-High, sovereign-cloud, OT, and classified research.
What's Included
- Everything in Enterprise, self-contained, zero phone-home.
- On-prem model serving, FIPS 140-3 crypto, TPM-attested boot.
- Threat-intel updates over a one-way air-gap channel (SIPRNET/WSUS pattern).
| Capability | AISE | Enterprise | NonPublic |
|---|---|---|---|
| Endpoint EXE client | ● Standard | ● Standard | ✕ Not available |
| SASE plug-in mode | ● Standard | ● Standard | ✕ Not available |
| Full AI client coverage | ● Standard | ● Standard | ● Standard |
| Edge appliance + on-board agents | ● Standard | ● Standard | ● Standard |
| SPM · ASM · TSM posture | ● Standard | ● Standard | ● Standard |
| Risk management (NIST AI RMF / OWASP) | ● Standard | ● Standard | ● Standard |
| IAM for agents (Entra Agent365) | ● Standard | ● Standard | ● Standard |
| Court-quality signed audit | ● Standard | ● Standard | ● Standard |
| Inline blocking | Optional | ● Standard | ● Standard |
| Attach to customer SASE / firewall | ✕ Not available | ● Standard | ✕ Not available |
| Customer-owned data plane | Optional | ● Standard | ● Standard |
| BYOK / customer key custody | ✕ Not available | ● Standard | ● Standard |
| Zero external dependencies | ✕ Not available | ✕ Not available | ● Standard |
| On-prem model serving | ✕ Not available | Optional | ● Standard |
| FIPS 140-3 validated crypto | ✕ Not available | Optional | ● Standard |
| Hardware-attested boot (TPM) | ✕ Not available | ✕ Not available | ● Standard |
| One-way CTI mirror (SIPRNET/WSUS) | ✕ Not available | ✕ Not available | ● Standard |
Frequently Asked Questions
What is Defiance AI™?
Defiance AI™ is an AI security platform that inspects every AI conversation on your network in real time and blocks prompt injection, jailbreaks, data exfiltration, and shadow AI usage. It sits between your users and any AI service, giving security teams visibility and control over AI traffic their firewalls cannot see.
How Is Defiance AI™ Different From A Firewall Or WAF?
Firewalls inspect network packets. WAFs inspect web requests. Neither one parses the natural language inside an AI conversation, which is where modern AI attacks live. Defiance AI™ runs alongside your existing security stack and handles the AI-specific traffic those tools were never designed to inspect.
What Problem Does Defiance AI™ Solve?
Defiance AI™ solves the AI security gap that traditional tools cannot cover. AI attacks arrive as ordinary sentences with no malicious code, no signatures, and no packet patterns to match. Firewalls, WAFs, and DLP tools miss them entirely. Defiance AI™ reads the intent behind each prompt and blocks the ones that put your data or systems at risk.
Can Defiance AI™ Detect Shadow AI?
Yes. Defiance AI™ identifies every AI service touching your network, sanctioned or not, from day one. Detection uses DNS queries and network metadata, so there are no agents to install, no browser extensions to push, and no dependency on employees reporting the tools they use.
Does Defiance AI™ Protect Self-Hosted AI Models?
Yes. Defiance AI™ protects self-hosted and on-premises open-source models with the same policy layer that protects cloud AI services. You can secure a mix of commercial LLMs, internal models, and third-party AI applications from a single console with consistent enforcement across all of them.
Will Defiance AI™ Work With Our AI Model?
Yes. Defiance AI™ is model-agnostic and works with OpenAI, Anthropic, Azure OpenAI, Google, AWS Bedrock, and self-hosted open-source models. One policy layer covers every provider you use today and any provider you switch to tomorrow, so your controls travel with you.
Will Defiance AI™ Slow Down Our AI Applications?
No. Passive deployment modes add zero latency because they inspect mirrored traffic. Inline enforcement is engineered to stay below the threshold users notice. Response times for end users remain the same before and after deployment, and there is no impact on model quality or output.
Does Defiance AI™ Store Our Data?
No. Defiance AI™ does not store, retain, or train on your data. The platform inspects AI traffic only, meaning the prompts and responses moving between your users and AI services. Everything else on your network is invisible to us. Inspection happens in real time and inspected content is not archived beyond what your policy requires for audit.
Does Defiance AI™ Require Decrypting Our Traffic?
Not initially. The first deployment stage uses only DNS queries and network metadata to discover AI usage. Decryption applies only when you enable deeper inspection, and only to AI-bound traffic. Everything else on your network stays encrypted and untouched.
How Does Defiance AI™ Handle AI Governance And Compliance?
Defiance AI™ maps to the PurpleSec AI Readiness Framework and to NIS2, GDPR, EU AI Act, DORA, ISO 27001, ISO 42001, OWASP LLM Top 10, MITRE ATLAS, PCI DSS, HIPAA, NIST CSF, NIST AI RMF, CMMC 2.0, and SOC 2 out of the box.
The platform detects your governance requirements automatically and enforces the matching policies without manual configuration. PurpleSec’s AI vCISO services are available to help you navigate AI governance strategy, audit prep, and regulatory change.
How Is Defiance AI™ Deployed?
Defiance AI™ deploys in three progressive stages. Stage one discovers AI usage passively with zero user impact. Stage two inspects prompts and responses and scores risk without blocking. Stage three enforces policy inline, blocks attacks, and rewrites unsafe responses.
Most customers start passive and enable enforcement once policy is tuned.
How Long Does Defiance AI™ Take To Deploy?
Defiance AI™ deploys in three progressive stages. Stage one discovers AI usage passively with zero user impact. Stage two inspects prompts and responses and scores risk without blocking. Stage three enforces policy inline, blocks attacks, and rewrites unsafe responses.
PurpleSec engineers work alongside your team through every stage, handling tuning, stage three escalations, and custom dashboards so the platform fits how you actually operate. Most customers start passive and enable enforcement once policy is tuned.
Who Manages Defiance AI™ After Deployment?
Your existing IT or security team can manage Defiance AI™ directly. The dashboard uses plain language rather than security jargon, and ongoing management typically takes minutes per week rather than dedicated headcount.
PurpleSec provides onboarding training, documentation, and ongoing support so your team runs the platform with confidence rather than guesswork.
For teams that prefer to outsource, PurpleSec also offers fully managed services covering monitoring, tuning, escalations, and reporting so you get the outcomes without staffing the operation.