NuDay AI provides security and governance software for organisations that want to put AI agents into production in regulated industries such as financial services, healthcare, government and defence, technology, human resources and life sciences. It tackles the risks that stall those projects: agents holding live credentials, tampered or injected tools, sensitive data sitting unencrypted in agent memory, and actions that nobody can later reconstruct for an auditor.
The platform wraps a zero-trust layer around existing agent infrastructure and uses cryptography rather than model behaviour to enforce its controls, an approach the company sums up as “Provable, not probabilistic.”
At A Glance
| Vendor | NuDay AI, a Galois company |
|---|---|
| Category | Zero-trust security and governance for agentic AI |
| Identities covered | AI agents, human users and service identities bound to agents |
| Deployment | Cloud-hosted, on-premises or hybrid, with air-gapped support |
| Agentic AI | Signed tools, on-behalf-of execution, encrypted agent memory, agent audit |
| Founded | 2025 |
| Market classification | Agentic AI Identity; also AI Governance & Compliance |
What NuDay AI Does
Guardrails that rely on a model behaving well can be talked around, and that uncertainty is hard to defend in front of a regulator. NuDay AI takes a different route: each control is enforced by cryptography and policy outside the model, so a check either passes or fails and the result can be shown later. Every tool call, data request and message between agents passes through one enforcement layer, which gives security and compliance teams a single point to approve deployments and to produce evidence for frameworks including SOC 2 Type II, HIPAA, SOX, PCI-DSS, NIST AI RMF, the EU AI Act and ISO 27001.
Products
Encryption Hub
The Encryption Hub protects the whole data surface an agent touches, which matters when agents store customer records, patient notes or internal knowledge in memory. Encryption is on by default.
- Encrypted Memory protects prompts, inputs and long-term history the agent saves and recalls
- Encrypted RAG encrypts the knowledge base and obfuscates the vectors as well
- Encrypted Shared Memory protects memories passed between agents and allows access to be fully revoked
- AES-GCM encryption, with ML-DSA (Dilithium) or classical DSA signatures chosen per tenant and ML-KEM for post-quantum key exchange
- Customer-held keys on SaaS or hybrid deployments, and keys kept in house on-premises
Policy & Access
Policy & Access solves the problem of agents carrying API keys and tokens that can leak or be abused. It keeps the agent’s identity separate from the user’s, so only one credential travels with a request, never both. For each action, the Policy Server checks policy and issues a short-lived credential scoped to that single action. Agents can act on behalf of a user or another agent, inherit permissions from the user’s organisational role, and pause for human approval when a request goes beyond their baseline rights. Behavioural drift detection adds a second check, flagging an agent that starts acting outside its normal pattern. For privileged work, short-lived credentials are checked out from the organisation’s existing PAM vault.
Signed Tools
Signed Tools defends against prompt injection and supply chain tampering at the tool layer. Every approved tool is hashed and signed with an enterprise private key, and the runtime checks the signature before anything executes. An injected tool fails the check and is blocked, and a hash mismatch reveals a compromised library that has altered the payload. Skills are bound cryptographically to the agent’s identity token, and verification is deterministic rather than based on scoring.
Agent Governance
Agent Governance gives risk and audit teams accountability for what agents do. Each agent is bound to a verified human or service identity, an approach described as “OIDC identity-bound agency”. Sensitive actions wait for a human to approve them, and the approval is logged before the action continues. Tracing records the full path from the user’s intent through the agent’s reasoning to the final action as an immutable forensic record.
How It Works
The platform can be cloud-hosted, on-premises or hybrid, and supports air-gapped environments. It works with models from OpenAI and Anthropic, and for sovereign AI the language model itself can run on-premises or air-gapped.
| Component | What it does |
|---|---|
| Enforcement layer | The “security mesh that wraps around your agentic AI infrastructure”, through which tool invocations, data access and agent-to-agent traffic flow. Traffic between agents is protected with mutual TLS on the same path as tool calls. |
| Policy Server | Sits between agents and the systems they call. It holds the secrets, runs each call on the agent’s behalf and returns only the data requested. |
| Encryption Hub | Sits under agent memory and retrieval stores, using per-checkpoint keys and an HMAC integrity check that runs before any data is decrypted. |
| Control plane | A central enterprise console with granular role-based access control, policy managed as infrastructure as code and human approval workflows. |
Identities And Systems Covered
| AI agents | Each agent has its own identity, distinct from the user it serves, and acts through on-behalf-of execution. |
|---|---|
| Human users | Agents are bound to a verified person, and their permissions follow that person’s role. |
| Service identities | Agents can be bound to a service identity instead of a person. |
| Secrets and credentials | API keys, tokens and privileged credentials, held centrally so agents never receive them. |
| Tools and skills | Signed tools and skills tied to the agent’s identity. |
| Agent data | Memory, RAG knowledge bases and shared context. |
Integrations
| Area | Integrations |
|---|---|
| Directory sync | LDAP, Active Directory |
| Approvals | Slack, Microsoft Teams, email |
| Audit and telemetry | OpenTelemetry |
Company Background
| Founded | 2025, when NuDay AI launched in stealth mode as a spinoff from Galois. |
|---|---|
| Ownership | A Galois company, built on Galois cryptography. |
| Team | Security architects, cryptography experts and AI researchers. |
Note: This overview is compiled from NuDay AI’s publicly available information as of 3 November 2026. It is provided for information only. NHI Mgmt Group does not endorse any vendor, and nothing on this page represents the opinions of NHI Mgmt Group.