Disclosure at collection. A business must inform consumers, at or before the point of collection, of the categories of personal information collected and the purposes for use.
Pre-compliance intelligence For government & policy teams
Read the policy once.Trust the Brief for years.
Civys reads legal and policy text and returns a structured brief: every claim tied to its source, every gap and uncertainty flagged, so nothing gets invented. Each brief is saved as a permanent record and routed through your team to review, edit, and approve. The speed of a chatbot, without the guessing, the forgetting, or going it alone.
Aligned to the NIST AI Risk Management Framework
Reading California CCPA
Opt-out of sale. If a consumer directs a business that sells personal information not to sell it, the business must comply.
How quickly must a business respond to a deletion request?
Reason: the text is silent. It sets no response deadline, so Civys declines to infer one.
Built to read dense legal & policy text
California CCPA
EU GDPR
Title IX
Clean Air Act §112
Local Ordinance 2024-18
HIPAA Privacy Rule
Dodd-Frank §1502
State Procurement Code
Zoning Variance Bylaw
Public Records Act
§01
How it works
From a wall of text to a record of work.
Faced with sixty pages, you could paste them into a chatbot. But a chatbot can invent details, it forgets the answer it gave you last week, and it leaves your team nothing to check. Civys runs the same text through a workflow built to prevent all three.
-
Dense text
Paste a bill, statute, or policy.
-
Nothing invented
Civys reads it
Every claim cited, every gap flagged.
-
A structured brief
What it says, what’s missing, what’s unclear.
-
Done together
Your team reviews
Edit, then a named person approves.
-
Never lost
Stored for good
Permanent, versioned, searchable.
Nothing invented. Nothing lost. Nothing done alone.
STEP 01
Paste the text
Drop in a bill, statute, ordinance, or internal policy. Before any of it leaves your computer, Civys checks the text for personal information right in your browser, so nothing sensitive gets sent out by accident.
Private on your device
STEP 02
Civys reads it
Civys works through the text point by point. For everything it tells you, it shows where in the document that came from and how sure it is. Not a loose summary you have to take on faith.
Same method every time
STEP 03
A person signs off
Every brief starts as a draft. Before it counts as final, a named person on your team confirms they’ve checked it against the original text, and that sign-off is saved with their name and the date.
Signed by a person
STEP 04
You keep the record
You end up with a finished brief that records what it was based on, who reviewed it, and how it was made. Download it as a PDF for a meeting, or as a data file that plugs into your own systems.
PDF & data exportWhat’s inside a Brief
SB-1042 · Automated Decision Systems in Public Agencies
ApprovedInventory of systems. Each agency must publish an inventory of automated decision systems in use, updated annually.
Impact assessment. If a system affects eligibility for public benefits, the agency must complete an impact assessment before deployment.
Does the inventory requirement cover systems built before this act?
Reason: “in use” is undefined. The text neither includes nor excludes legacy systems, so Civys declines to infer.
SHA-256 a3f2…9c41
Model claude-sonnet-5
Prompts v2.0
Citations 14/14 claims tagged
A person signs off
A brief is only marked “Approved” once a named reviewer has checked it against the original. Every change in status is logged, and the log can’t be edited after the fact.
What’s required, and how firm
Every requirement is sorted into mandatory, discretionary, or conditional, based on the language the text itself uses, with a note on how clear-cut each call is.
Where each point comes from
Each point shows how close it is to the original, whether it’s a direct quote, a paraphrase, or a reasoned read, and links to the exact section, so you can check it in seconds instead of rereading pages.
What the text leaves unclear
Vague terms and open questions are listed out, not glossed over. This is the part that keeps a brief from sounding more certain than the law actually is.
The full paper trail
A record of exactly what went into the brief: the original text it read, the AI model, the method version, and how much of the brief is backed by a citation. Any brief can be traced straight back to its source.
§02
In practice
Built for the work policy offices actually do.
Three places a brief earns its keep, from the Monday bill drop to the audit two years later.
Bill analysis
Session drops forty bills on Monday.
See what each one actually asks of you before the hearing, not after. Civys lays out what’s required and what’s optional, who’s covered and who’s exempt, and flags what a bill leaves unsaid instead of guessing at it, so you walk in briefing on what the text really says, with no risk of a chatbot’s invented details.
For policy analysts & legislative staff
Mandatory
Discretionary
Conditional
Stakeholder briefing
Everyone asks: what does it mean for us?
Turn a brief into talking points you can stand behind. Every line points back to the exact spot in the text it came from, so when someone pushes back, you show them the source on the spot, backed by a saved record you can reopen a year from now instead of a chat that’s long gone.
For chiefs of staff & department heads
DIRECT
PARAPHRASE
INTERPRETATION
Compliance tracking
The auditor asks: “how did you determine that?”
Every requirement gets its own entry with a status and a full history. Your team’s read is written down, sourced, and signed off before anyone asks, a permanent, searchable record instead of a folder of marked-up PDFs.
For compliance officers & counsel
Draft
In review
Approved
§03
Why Civys
A chatbot leaves you with three problems. A brief removes all three.
It can fabricate details. It keeps no record you can return to. And it gives your team no way to review what it produced. Every part of a Civys brief answers one of those three.
100%
of claims carry a source tag, enforced by the prompt suite and measured on the output
4
trace levels (DIRECT, PARAPHRASE, INFERENCE, INTERPRETATION) so you know how far each claim is from the text
3
obligation classes (mandatory, discretionary, conditional) driven by the text’s own signal words
0
invented conclusions. Where the text is silent, the brief says “no conclusion available”
Built to say “no conclusion available.”
Chatbots are rewarded for sounding sure, which is exactly how they end up inventing things. Civys is built the opposite way: when the text doesn’t answer a question, the brief records the question, the reason, and nothing else. It flags what’s uncertain instead of papering over it, so you never mistake a guess for the law.
Are contractors subject to the reporting requirement?
No conclusion available: the definition of “covered entity” in §2 does not mention contractors, and no other section resolves it.
No Brief approves itself.
Your team edits the draft together, then a named reviewer signs off. Nothing becomes a record without a person’s explicit approval, stored with their name and the time:
“I have reviewed this brief against the source policy text…”
Pinned to the exact text.
Unlike a chat history that scrolls away, every brief is saved for good and stamped with a fingerprint of the exact text it read, so months later you can prove which version of a bill your interpretation was based on, even after the bill is amended twice.
SHA-256 a3f2c17e884b…06d9c41
Every claim shows its distance from the source.
A direct quote and an interpretation are different animals, and a brief never lets them wear the same clothes. You can see at a glance how close each line is to the actual text.
DIRECTPARAPHRASEINFERENCEINTERPRETATION
An audit trail, not a chat log.
Each brief carries an ID, a status, and a full history of every change (submitted, approved, reverted), each one stamped with who did it and when. Your team’s work is preserved, not scattered across inboxes and chat windows.
DraftIn reviewApproved
Disclosure built in
An AI factsheet on every Brief.
Model, prompt version, input hash, metrics, and review status, embedded in each brief and its exports, structured after the GovAI Coalition template and mapped to the NIST AI Risk Management Framework.
NIST AI RMFGovAI FactSheetPer-brief metrics
§04
Built for the standard
Ready for the questions a public-sector buyer asks.
Civys is built to the frameworks that govern responsible AI in government: a per-brief AI factsheet structured after the GovAI Coalition template, mapped to the NIST AI Risk Management Framework, with human oversight and data handling documented rather than assumed.
Governance-first. Client-side PII screening, no model training on inputs, and user-deletable records.
Human-attested. Every approved brief carries a named reviewer’s explicit sign-off, logged with identity and timestamp.
Fully traceable. Any output can be traced to the exact model, prompt version, and source text that produced it.
§05
Get access
Bring the next bill.
Create an account, paste the text you’re already expected to understand, and read your first brief in minutes. Judge it the way you’d judge a colleague’s memo: against the source.
- Minutes to first brief. No setup, no integration required to start.
- Your records stay yours. briefs live in your account, exportable and deletable at any time.
- Honest by design. If a brief can’t support a conclusion, it will tell you exactly that.
Access your briefs
Sign in
Your briefs and audit trail, exactly where you left them.
or
If you don't have an account yet, continuing creates one and confirms you are 18+ and agree to the Terms and Privacy policy.
New to Civys? Create an account