Owner approval
One research note states a development rule: the owner approves a change before it happens. It is not a study, a customer result, or a performance claim.
In development
United States AI Security and Defense Research Center
A privately operated division and registered DBA of Ultrec AI Inc
Not affiliated with or endorsed by any government agency
The Center develops AI-assisted cybersecurity software for people and organizations who need a plain explanation and an owner-controlled next step. No product listed here is released. Nothing on this site can be bought, installed, or switched on.
Where the work stands
-
Private
A privately operated division and registered DBA of Ultrec AI Inc. It is a DBA, not a separate company, and not a government body.
-
Not released
Five AI Defense Agents and five Softwares are listed below. None is released. Trojan Eliminator Agent, an AI Defense Agent, is approved for development. It is not built, tested, or released.
-
No download
There is no cart, license, price, or installer. Questions go to the addresses on the contact page. This website does not store those messages.
Trojan Eliminator Agent
Trojan Eliminator Agent is a working name, and it is not trademark-cleared. It is approved for development. It is not built, tested, or released. The intended role is to explain a finding an established scanner already made, then wait until the owner approves a specific change. It does not guarantee malware removal.
AI Defense Agents
These five are AI Defense Agents, not Softwares. Working names are not trademark-cleared. None is released, and none is a download. Open AI Defense Agents.
Approved for developmentTrojan Eliminator AgentA proposed Windows-first assistant. It would explain a finding already made by an established antivirus or EDR product, then wait while the owner approves any change.- Not releasedSecret SentryWould look through repositories, logs, and process environments for leaked credentials.
- Not releasedPatch PilotWould inventory Windows and macOS applications and missing updates, then propose a patch plan for the owner to approve.
- Not releasedExposure CheckWould check open ports and weak settings only on a domain or devices the owner is explicitly allowed to test.
- Not releasedSecurity Briefing DeskWould summarize advisories that already name their sources, in language a non-expert can use.
Softwares
These five are Softwares, not AI Defense Agents. Working names are not trademark-cleared. None is released, and none is a download. Open Softwares.
- Not releasedPhish InspectorWould explain a suspicious message, link, or attachment using reputation lookups and header analysis. It would not delete mail on its own.
- Not releasedAgent FirewallWould block an AI agent from a risky tool, file path, or outbound domain unless a person approves that step.
- Not releasedPrompt-Injection ShieldWould screen a document or page before it is given to an AI agent and flag text that looks like an injected instruction.
- Not releasedIncident NotebookWould help a small office keep an incident timeline, an evidence log, and a plain-language report.
- Not releasedRed-Team RangeWould exercise only the owner's own AI agents inside a sandbox and score the defenses. It is not for attacking anyone else's systems.
Both catalogs remain together at the earlier address, with the two lists kept separate.