Responsible AI risk intelligence

Find true vertical on AI risk.

An honest read on AI risk, built from evidence, not a questionnaire. A grade, the gap between claimed and proven, and the questions worth asking next.

Three free briefs at launch. No login. Your inputs are never stored.

You are on the list.

The Plumb Grade, start to finish

Plumbline · live read
Give it a subject
PDFAI_Governance_Policy.pdf
XLSvendor_risk_checklist.xlsx
Provenancetraceable to a credible source
Qualitystrength of each piece on its own
Consistencypolicy, product, and public story agree
Completenesshow much of what should exist does
Responsible AI risk brief

-

Plumb Grade
-
-
-
Q1

-

Illustrative. Composite subjects, not real companies.

Who this is for

You already think in evidence.

You put AI into real decisions

Insurers, banks, hospitals, large employers. You moved AI into choices that affect people, and you know principles are not proof. Find your own gap before a regulator or a reporter does.

TriggersBefore deploying a new model. After a vendor update. Before a regulatory review. Every quarter as a baseline.
You depend on someone else's AI

Buyers, acquirers, investors. Every vendor ships AI now, and the newest tools carry risk no one has mapped, sometimes not even the team that built them.

TriggersBefore you sign. At renewal. The week a vendor ships a major AI feature.
You answer for it

Boards, risk committees, CIOs. When someone asks whether you are exposed on AI, you want a defensible answer, not a feeling. Walk in with the grade, the gap, and the questions already framed.

TriggersBefore every committee cycle. Before any regulator or insurer conversation.
The new risk
The fastest AI tools carry attack surfaces their own builders have not mapped. When it breaks, the company pays.

Plumbline reads whether a vendor can show its security or only assert it. The gap between those two is where the next incident lives.

How it works

From a name to a brief you can act on.

01

Name the subject

A company, a URL, or pasted policy text. There is no questionnaire. That is the whole idea.

openai.com/securityguidewire.com/productsSEC 10-K risk factorsvendor_security_response.pdf
02

Plumbline reads the evidence

It gathers what is public and reads anything you provide directly: policies, security questionnaires, vendor documentation. It weighs the evidence for provenance, quality, consistency, and completeness, and separates what is shown from what is only claimed.

03

You get the brief

The Plumb Grade, the assertion gap, the top risk driver, a confidence level, and the questions worth asking next.

04

Take it to the decision

Export the brief as a card and bring it to a committee, a vendor conversation, or your own planning. It reads clearly to people who do not live in AI.

The Plumb Grade

One letter. How far a company leans on claims it cannot show.

True verticalFull lean
A
Claims fully demonstrated
B
Key claims substantiated
C
Partial evidence, major gaps
D
Asserted, not shown
F
Opaque, no meaningful disclosure
Cloud platform
B
Confidence High
Outlook Stable
Health insurer
C
Confidence Moderate
Outlook Deteriorating
AI vendor
D
Confidence Low
Outlook Developing
Pricing

A brief costs less than the meeting where you need it.

01

Your vendor stack changes every quarter

Every new SaaS tool, every new model API, every new integration is a new assessment trigger.

02

The vendors themselves change

A vendor graded B in January could be D by September. An acquisition, a new model, an enforcement action. The outlook is the signal to come back.

03

The rules are moving

EU AI Act enforcement is live. US state legislation is accelerating. Every new regulation is a re-assessment trigger.

So the model is simple. Start free. Pay per brief when you need one. Subscribe when you own the portfolio.

Free
$0

You heard about it. Try it.

  • 3 briefs, no account needed
  • Plumb Grade (A through F)
  • Assertion gap and top risk driver
  • One key question
  • Full evidence breakdown
  • Export and share
  • Saved history
Run your first brief
On Demand
$29per brief

You have a vendor to assess before you sign.

  • Pay per read, no subscription
  • Plumb Grade with full evidence breakdown
  • All four dimensions scored
  • Full question set
  • Confidence level and outlook
  • Export as a shareable card
  • Saved brief history
Run a full brief

No subscription. Buy one when you need one.

Continuous
$249per month

You own the portfolio and answer for it.

  • Unlimited briefs
  • Everything in On Demand
  • Saved vendor portfolio
  • Re-assessment alerts when vendors change
  • Concentration risk flags
  • Side-by-side comparisons
  • Team seats (up to 5)
Talk to us

Annual billing available.

The grade never changes based on what you pay. Only how much of the brief you see.

How it reads the evidence

Four principles a questionnaire cannot reach.

01

Evidence is weighted by what backs it

An audited control counts for more than a written policy, which counts for more than a public claim. Plumbline scores the strength of the proof, the way a credit analyst trusts a filing over a press release.

02

The assertion gap is the headline

The number that matters is the distance between what a company claims and what it can show. Auditors and short sellers live in that gap. For AI risk it is the finding.

03

Confidence is separate from the grade

A strong grade we cannot verify is a different conversation than one we can. We never blend the two. Low confidence at any level is something worth hearing out loud.

04

Outlook is a direction, not a level

The grade reads where the evidence stands today. The outlook reads where it is heading, and whether the proof is deepening or thinning.

The stance underneath all of it

Absence of evidence is evidence. We score opacity as risk. We never wave it through as neutral.

A rating is only as trustworthy as the incentive behind it.

In 2008 the agencies that stamped mortgage bonds AAA were paid by the people issuing them. The conflict was built in, and it was fatal.

Plumbline runs the other way. It will never raise a grade for the company being graded. The people who pay are the people who need the truth. A leader doing diligence, an insurer, a buyer assessing risk.

That one choice is why a Plumbline read can be trusted to be uncomfortable.

Why this exists

I spent my career where trust has to be earned, not claimed. Capital markets, custody, the software banks run on, insurance. The job never changed. Separate what is verified from what is asserted, and price the difference. AI governance has that exact problem, and nothing built to measure it. So I built one.

Nithila Jeyakumar, Founder. MIT, System Design and Management
Built honest

What happens to what you give it.

Keys stay on the server

We call the model from the server. No credential ever reaches your browser.

URLs are fetched with care

Public links only, over HTTPS, with private and metadata addresses blocked, plus timeouts and size caps.

Your input is data, never instructions

Everything you paste is treated as untrusted content, so it cannot redirect the assessment.

Nothing is kept

We do not log what you send, and we do not store your briefs. Every read is stateless.

We hold ourselves to the bar we measure others against.

Plumbline is in early access. It sharpens the questions worth asking. It does not replace counsel, an audit, or a regulator, and it does not pretend to.

Early access

See what your evidence actually says.

Free at launch. On Demand and Continuous coming soon.

You are on the list.