The Deflected Platform, Explained

Executive summary

Deflected is a single platform that secures the AI layer of a modern enterprise — the models, prompts, agents, and data pipelines behind every AI feature you ship or adopt. It combines always-on software products with expert services, and encrypts everything it touches with post-quantum cryptography that is designed to withstand both today's attackers and tomorrow's quantum computers.

This page explains the whole platform in one place, written for the people who have to approve, budget for, and stand behind AI security decisions: CISOs, heads of AI, CTOs, risk and compliance leaders, and the boards they report to. Every technical term is defined the first time it appears. Where a capability has its own dedicated page, we link to it so you can go deeper.

What the "AI layer" actually is

Traditional security is organized around familiar layers: the network, the endpoint, the application, the identity system. Artificial intelligence introduces a new layer that sits on top of all of them. The AI layer is everything that makes an AI feature work:

• Models — the large language models (LLMs) and other machine-learning systems that generate text, decisions, code, or predictions.
• Prompts — the instructions and context sent to a model, often assembled from user input and internal data on the fly.
• Agents — AI systems granted tools and autonomy to take actions: calling APIs, querying databases, sending messages, or executing code.
• Retrieval pipelines — systems that fetch documents and data (for example from a vector database) and inject them into prompts, commonly called retrieval-augmented generation, or RAG.
• The data — the proprietary information, customer records, and model weights that all of the above touch.

Every one of these components is a place where something can go wrong in a way that older tools do not understand. That is the gap Deflected was built to close.

Why the AI layer needs its own security

The uncomfortable truth for most enterprises is that the tools protecting their AI today were designed for a different problem. A web application firewall inspects HTTP traffic for malicious code. It has no concept of a sentence that politely instructs a model to ignore its own safety rules. That is the essence of the AI-era threat model: the attacks are written in natural language, and they target behavior rather than infrastructure.

Prompt injection: the defining AI vulnerability

When an application inserts untrusted input — from a user, or from a document it retrieved — into a prompt, an attacker can hide instructions inside that input to hijack the model. This is called prompt injection, and it is to AI what SQL injection was to databases: the single most common and consequential vulnerability. It can be used to exfiltrate data from the model's context, manipulate tool calls, or bypass guardrails entirely.

Data leaving through the model's output

An AI system can leak sensitive information simply by generating it. A model with access to customer records or internal documents can be coaxed — or can accidentally be led — into revealing them in its response. Conventional data-loss prevention watches files and network traffic; it does not read model output for regulated data escaping in plain language.

Impersonation and synthetic media

Generative AI has made convincing voice clones and deepfakes cheap. Business email compromise and executive impersonation — long-standing corporate fraud vectors — are now supercharged by an attacker who can produce a CFO's voice on demand.

The quantum horizon

Finally, there is a threat that is invisible today but already in motion: adversaries are capturing encrypted data now and storing it to decrypt later, once quantum computers can break today's cryptography. This is known as harvest now, decrypt later, and it means any data that will still be sensitive in five or ten years is already exposed unless it is encrypted with post-quantum algorithms.

The platform, layer by layer

Deflected organizes its coverage into two categories. Products are always-on software capabilities, billed monthly, that run continuously in your environment. Services are expert engagements, scoped and quoted per environment, where our specialists do focused work with your team. Together they cover the full lifecycle of AI risk — from the code you ship to the board report you present.

Products — always-on protection

Services — expert engagements

Quantum-grade encryption, explained

Every byte that flows through Deflected is protected with post-quantum cryptography — encryption based on algorithms that resist attacks from both classical and quantum computers. This is not a premium add-on; it is the default across the platform. Here is why it matters and what we use.

Most of today's internet security relies on math problems that are hard for classical computers but that a large quantum computer could solve quickly using Shor's algorithm. When that day comes, data protected only by classical public-key cryptography becomes readable. Because adversaries are already harvesting encrypted data to decrypt later, the safe posture for any long-lived data is to encrypt it with post-quantum algorithms today. Deflected uses the standards finalized by the U.S. National Institute of Standards and Technology (NIST):

• ML-KEM-1024 (formerly CRYSTALS-Kyber, NIST FIPS 203) for key encapsulation — securely exchanging keys at a 256-bit quantum security level.
• ML-DSA-87 (NIST FIPS 204) and SLH-DSA (NIST FIPS 205) for digital signatures that remain unforgeable in a post-quantum world.
• Hybrid X25519 + ML-KEM key exchange, which runs a proven classical algorithm alongside the post-quantum one, so you are protected even if either scheme is ever weakened.
• AES-256-GCM for symmetric encryption of data at rest and in transit.

The practical benefit for an enterprise is a promise very few competitors can make: the data your customers entrust to you is protected not only against the threats of today, but against the cryptographic threats of the coming decade.

Governance & compliance

Security that cannot be demonstrated is incomplete. Regulators, auditors, and enterprise buyers increasingly ask organizations to prove that their AI is governed responsibly. Deflected maps its controls and evidence to the frameworks that matter most:

• NIST AI Risk Management Framework (AI RMF) — the leading voluntary framework for identifying and managing AI risk across the model lifecycle.
• EU AI Act — the European Union's risk-based regulation of AI systems, with real obligations for high-risk use cases.
• SOC 2 — the trust-services criteria enterprise procurement teams rely on to evaluate a vendor's security posture.

The outcome is an AI program that is audit-ready, not just secure: framework mappings, implemented controls, and the evidence and reporting to back them up. For a corporate team, this is often the difference between a stalled deal and a signed one, or between a clean audit and a painful one.

How deployment works

Deflected is built to run in production without becoming a project that never ends. The path is deliberately short:

1. Map — we identify where AI touches your business: the models, features, agents, and data flows that make up your AI layer.
2. Integrate — always-on products such as Prompt Firewall are placed inline in the request path with sub-second latency and safe fallbacks that never break your application.
3. Tune — detection is calibrated to your specific applications and policies, so you get signal, not noise.
4. Operate — you get a readable dashboard, alerting on what matters, an immutable audit log of every decision, and expert services on standing retainer for the moments that need a human.

Who it's for

Deflected is built for organizations deploying AI where the stakes are real — where a breach means regulatory exposure, lost customer trust, or material financial harm. That includes:

• Financial services — protecting customer data, models, and approval workflows against fraud and exfiltration.
• Healthcare — keeping regulated patient data sealed against both today's attackers and the quantum horizon.
• SaaS companies — securing AI features so they can be sold into demanding enterprise and regulated markets.
• Government & public sector — meeting high assurance requirements for AI and cryptography.
• Legal & professional services — safeguarding privileged and confidential information handled by AI systems.

The common thread is leadership that understands AI is now core infrastructure — and that securing it can no longer be deferred.

Frequently asked questions