Latest Treasury Secretary Scott Bessent on the rogue-agent breach: “It is the humans who are responsible, not the AI.” Why that lands on managers →
Forge Orbital  ·  Independent AI decision accountability

Forging the future of AI accountability.

Use AI in your business. Keep control of what it does.

Forge is ready for your business.
Independent businesses. Mid-market firms. Global enterprises. Any industry.
AI governance, evaluation and optimization—built into the work your business depends on.

DoD CDAO Tradewinds Solutions Marketplace Assessed as Awardable badge
Assessed as Awardable DoD CDAO Tradewinds Solutions Marketplace
NVIDIA Inception Program badge
NVIDIA Inception Program member
Cloud Security Alliance AI Trustworthy Pledge 2026 badge
Scroll
Why now

You can’t grade
your own homework.

You can’t just throw your AI into another AI and hope it audits it. Asking Gemini to check ChatGPT can help find mistakes. It does not, by itself, prove the answer was right or the action was authorized.

ANOTHER OPINION

“The other AI said
it looked fine.”

Two models can agree and still be wrong. Agreement alone doesn’t show what was checked.

EVIDENCE YOU CAN EXAMINE

“Here is what we checked.
Here is who approved it.”

The source material, the rule, the result and the approval—connected to the action, so a reviewer can examine the basis for it.

Regulators are naming the problem

AI does the work.
Your business still answers for it.

FINANCIAL SERVICES / FINRA

Existing supervision rules still apply when member firms use generative AI—including third-party tools. Model accuracy, data integrity and oversight remain part of the job.

Read FINRA’s guidance ↗
TAX PRACTICES / IRS

The IRS tells practitioners to verify AI-generated facts, citations and calculations, and document their AI use and checks. An AI answer does not replace professional judgment.

Read the IRS guidance ↗
$100M

The cost of failed oversight.

In June 2026, the SEC ordered Western Asset Management to pay a $100 million penalty over trade-allocation and supervision failures. The case included missing timely allocation instructions and failures to follow documentation and approval policies.

This was not an AI case. It shows why controls and contemporaneous evidence matter—not that any particular AI tool is required.

Read the SEC order ↗

Whether you’re reviewing a client’s tax advice, an insurance recommendation or an investment decision, “the AI said so” is not the explanation you want to rely on. Keep the evidence. Make approval clear. Be ready to show your work.

Requirements vary by activity and jurisdiction. These sources are not endorsements of Forge. Forge supports review and accountability; it does not guarantee accuracy or compliance.

The Forge solution was assessed as Awardable on the U.S. Department of Defense CDAO Tradewinds Solutions Marketplace. This is an assessment, not a contract award or certification.

Here’s how Forge helps ↓
What we do

What does Forge Orbital
actually do?

Forge Orbital is the accountability layer for AI agents. Governance is where it starts: before a connected AI agent takes an action, Forge checks it against your rules and returns a result—allow, hold or block. It also creates a signed decision record that can be checked and replayed later. Your system enforces the result; your people retain authority.

BUILT INTO THE WORK

Not another
dashboard to watch.

Keep ChatGPT, Claude, Gemini and your own systems. Connect Forge to check the work against your rules and route exceptions to the people and tools you already use.

From client emails to bank model changes and hedge-fund workflows. The depth of the checks depends on the workflow and how we integrate.

Customer-hosted, on-premises or in your private cloud. Forge does not train product models on your data or learn across customers.

No exception flagged means none was found by the checks that ran—not that everything is safe. Connection and check failures need their own escalation path.

Set the rules before AI acts

Define what your AI agent is allowed to do, what needs a person’s approval and what should be stopped. Forge Orbital checks the proposed action against those rules.

Explain a decision later

Keep a signed record of the checked decision and the rules used. If a customer, auditor or leadership team asks what happened, you have evidence to review and replay.

Review quality and cost

Forge Orbital’s evaluator and optimizer help you assess agent performance and cost alongside governance, so you can improve a workflow without dropping accountability.

Fit it into your existing systems

Forge Orbital runs on-premises and integrates with your existing workflows. It is not another chatbot your staff has to prompt for governance.

AllowThe checked action meets your rules.

HoldPause for the review your workflow requires.

BlockStop an action that fails the rules.

For your technical team: what each capability does

Governance: returns a disposition for a connected action. Your application enforces it; people retain authority.

Evaluation: compares agreed cases and evidence. It needs representative inputs and expected outcomes.

Optimization: recommends candidates using supplied usage, prices and evaluation evidence. It does not automatically switch models or guarantee savings.

See it in context

A useful answer isn’t
permission to act.

One concrete question: does this action have the required approval? Run it without approval, then include approval and rerun. The result changes from held to cleared under that rule.

ILLUSTRATIVE · NOT A LIVE ENGINE CALL
AI-DRAFTED CLIENT EMAIL

“We can confirm your refund.”

The agent has drafted a promise to a customer. Before it sends anything, your approval rule applies.

YOUR BUSINESS RULE

A refund promise needs the account owner’s approval.

1AI proposes
2Forge checks
3Your system responds
Ready to check.

Press Run to follow this action from proposal to review.

Ready · 3 steps
What can your reviewer reconstruct later?

The proposed action, evidence supplied, policy version, check result and recorded approval. For an order, that can include the strategy mandate and risk inputs provided to Forge. For a model change, it can include the model versions and evaluation evidence.

Replay reconstructs the recorded Forge decision from its inputs and configuration. Verification checks that the proof has not been altered. Neither guarantees a model will repeat its answer, proves every input true or replaces the customer’s regulatory obligations.

Forge returns a check result. Your connected system enforces it, and your people keep authority. This browser illustration demonstrates one approval rule—not a live Forge engine call, signature verification or a connected customer system. Production checks depend on the evidence, rules and integration you configure.

See the business fit →Explore the integration →
Your team

People who know the stakes.
And speak your language.

Experience in financial services, national security and enterprise technology. A direct conversation with the people doing the work.

Financial servicesEnterprise technologyNational securitySystems engineering
JV

Joseph Varga

Founder and Chief Executive Officer

Built AI/ML systems for graphene-enhanced ballistic armor delivered to DEVCOM, then spent three years leading go-to-market in defense technology, before founding Forge.

Background ↗

BI

Brittany Istenes

Chief Trust & Ecosystem Officer

Brittany brings 15+ years of experience across enterprise technology, compliance, open source and governance, including a decade within Fortune 20 organizations. Her experience includes serving as an Open-Source Program Office Strategist at Fannie Mae.

Background ↗

RP

Rob Parkhurst

Chief Strategy Officer

Nine years as an intelligence specialist at Naval Special Warfare Command, across geospatial and financial intelligence.

Background ↗

DS

Declan Sheehy

Senior Advisor, Financial Services

Declan brings 25 years in regulated financial services, including 18 years leading operations and technology at HSBC Alternative Investments as the business grew from $200 million to $31.6 billion. He built the technology, oversight and vendor governance supporting that growth, with zero FCA breaches during his eighteen years at HSBC Alternative Investments. Now he brings that experience to Forge, helping connect AI accountability to the needs of financial services buyers in the UK and Europe.

Background ↗

More about Declan

He has since held COO and Chief Revenue Officer roles in fintech and regtech.

He holds the IAPP AI Governance Professional (AIGP) certification and is the author of Disruptive Innovation in Financial Services.

KH

Kyle Helseth

Strategic Advisor

Four years building federal go-to-market at a Y Combinator-backed, Series C technology company, most recently as a regional vice president serving national-security and defense customers.

Background ↗

WH

Will Hopkins

Integration Engineering Advisor

Spacecraft systems engineer with experience in satellite architecture and technical integration, including work involving the United Arab Emirates Space Agency and ITAR-controlled requirements.

Background ↗

Backgrounds describe individual experience, not customer relationships or endorsements of Forge Orbital.

In the Media

Joseph Varga on Insurance Town

A conversation with Heath Shearon about AI accountability, agentic AI and human oversight in insurance.

Any business. Global enterprise. Any industry.

If AI does work
you depend on, you need Forge.

Let’s talk about where it fits in your business.