The AI you can tell anything. Because it keeps no record.
- →Never written to disk. No chat logs. No transcripts. Nothing to produce in discovery.
- →Self-hosted model. Your prompts never touch OpenAI, Google, or Anthropic.
- →No name required. Numbered accounts. Nothing on file that ties a session to you.
- →Federal & 50-state statute search — cited to the section, and never logged.
$0.00 today · 2 hours free · then from $49.99/mo · Cancel anytime
The questions you need to ask are the ones you can't put in writing.
You won't ask your spouse. You won't pay $400/hr to ask a lawyer. So you ask ChatGPT — and it's now on a server, timestamped, tied to your account.
- HRAn employee is claiming harassment. How do I handle this without making it worse?
- ClaimsA customer says they slipped in my store. My video shows they didn't.
- LitigationI'm being sued. What's my exposure before I hire anyone?
- ReputationA fake review accuses my staff of something that never happened.
- ExitA former employee is making accusations that aren't true.
Real or fraudulent, every one is sensitive. And the place you typed it is the easiest evidence anyone will ever pull.
A federal judge ruled consumer AI chats aren't privileged.
Judge Jed Rakoff held that a defendant's chats with a free-tier AI tool were protected by neither attorney-client privilege nor work product — because the AI isn't a lawyer, the platform's own privacy policy defeats any expectation of confidentiality, and sharing the output with counsel afterward didn't cure it.
United States v. Heppner, No. 25-cr-00503 (S.D.N.Y. Feb. 17, 2026) — first ruling of its kind"Delete" isn't delete
OpenAI retains deleted chats where legally required — and a preservation order can hold them indefinitely.
No warrant needed
Their policy permits disclosure to authorities on a good-faith belief. You aren't notified.
Consumer tiers train on you
Free and $20/mo plans use your inputs by default and carry no contractual confidentiality.
If a leak would cost you, this is for you.
We can't leak what we never had.
This isn't a privacy policy. It's a build decision — made four times.
Encrypted connection
Your messages travel to us over an encrypted link. Locked in transit.
Numbered account
No name or email required. Add one only for recovery — we hash it if you do.
Our own model, our own servers
Self-hosted. Your prompts are never sent to a third-party lab.
Never stored
Held in volatile memory for the length of the reply, then gone. Never written to disk. Never logged.
If we're served, there's nothing to hand over. No logs, no stored chats, no identity on file. You can't produce what was never kept.
What they keep. What we don't.
| Consumer AI | IronBrief | |
|---|---|---|
| Your chat about a lawsuit | Logged, timestamped, discoverable | Never written down |
| "Deleted" conversations | Held 30+ days; indefinitely under a hold | Nothing to retain |
| Your identity | Email, payment and IP tied to every chat | No name or email required |
| Legal requests | Disclosure on "good faith belief" | Nothing on file to disclose |
| Training on your business | On by default for consumer tiers | Never — nothing retained to train on |
Know the actual law. Privately.
Ask in plain English. Get an answer grounded in primary law, cited to the exact section, scoped to your state. Nothing you search is logged.
Walk in informed
Understand your situation before you pay anyone to explain it.
Spend their rate on strategy
Not on getting counsel up to speed at $400/hr.
Zero query logging
Nothing you ask is retained — not the question, not the result.
Not a law firm. Not legal advice. A private research tool.
Frontier stacks vs what free tiers actually serve.
Two different IronBrief stacks — Basic (standard) and Deluxe (decensored) — both land in the public Artificial Analysis Intelligence Index frontier band. Compared below to the models most people get on free consumer plans, named so you can verify the board.
Intelligence Index
Public composite score. Higher = stronger general intelligence.
Scores ≈ current Artificial Analysis Intelligence Index for the named free-tier defaults (Gemini 3.7 Flash, Claude Sonnet 5, GPT-5.6 Luna) — not Fable / Opus / Sol max (paid caps, credits, or limited access). Basic = standard IronBrief weights · Deluxe = decensored IronBrief weights. Both track the open-frontier ~60 band; different stacks, same intelligence class. Index moves as releases ship.
Try free for 2 hours. Then pick your volume.
Every plan starts at $0.00 today with a free 2-hour trial. After that, self-hosted, encrypted in transit, zero-retention — the only difference is how much you can ask.
Starter
- ✓Encrypted & zero-retention
- ✓50-state legal search
- ✓Anonymous account
- ✓2 hours free · then monthly
Professional
- ✓Everything in Starter
- ✓3× monthly volume
- ✓Priority queue
- ✓2 hours free · then monthly
Scale
- ✓Everything in Professional
- ✓5M tokens every month
- ✓Priority capacity
- ✓2 hours free · then monthly
100% decensored. Zero refusals. Debanked on purpose.
Same private architecture — encrypted, zero-retention, nothing to subpoena — running an obliterated apex stack that answers what Claude, GPT, and Gemini are built to refuse. Card networks banned it. That’s the receipt.
Consumer AI caps your edge.
This is the edge.
Anthropic and OpenAI train the smartest models on earth — then wrap them in a compliance layer that refuses your hardest questions, hedges your sharpest ideas, and logs every word for discovery. That product is built for everyone. So it serves no one seriously.
Edge removes the muzzle. High-stakes strategy. Hostile diligence. Pricing wars. The conversations a corporate model is contractually forbidden to finish. Full answers. No apology. No paper trail.
- 100% decensored — no leash, no hedge
- Peak Index 148 — clears public flagships
- 25M tokens / mo — serious volume
- Zero retention — nothing to subpoena
- Debanked — cards cannot touch it
- Manual onboarding — not self-serve
- $1M+ company revenue — hard gate
We will turn you down. Edge is not a consumer upgrade. Company revenue must be over $1,000,000 — at or under $1M is declined without review. If you want a swipe-and-go chatbot, buy Starter. If you clear the gate — apply. Most inquiries don’t.
- Zero refusals — never says no
- Obliterated stack — apex intelligence
- Encrypted + zero-retention
- 50-state legal search included
- Anonymous architecture
- Human intake — 48h response
Who gets in: companies with over $1,000,000 in revenue. Below that we will not consider the application. Why debanked: fully unfiltered AI violates every card-network AUP — we kept the intelligence and cut the rails.
Request Edge access →Contact form only · company revenue required · no self-serve · no cards. Under $1M: don’t apply.
Self-serve: free 2-hour trial, then $49.99 / $99.99 / $149.99 per month at chat.ironbrief.ai. Edge is contact-only · $999/mo · crypto/wire · $1M+ company revenue required. Not legal advice. Not a law firm.
What you're actually getting.
Is IronBrief a lawyer or a law firm?
No. It's a private AI research tool. It doesn't give legal advice or represent you, and it isn't a substitute for an attorney. Nothing you type here is privileged — but unlike ChatGPT, nothing you type here is stored either.
How is this different from "temporary chat" on ChatGPT?
Temporary chat still sends your data to their servers, and it can still be retained under a legal hold. IronBrief runs a self-hosted model and never writes the conversation to disk. They promise not to keep it. We built it so there's nothing to keep.
What do people actually use it for?
Thinking through an employee complaint. Sizing up exposure on a claim. Drafting a response to a bad review. Researching your state's law on a dispute. Prepping questions before an attorney meeting.
If I lose my password, can you recover my history?
No — that's the point. You get a one-time recovery code at signup. Keep it somewhere safe. Real privacy means we can't get in either.
Stop leaving a paper trail.
The questions that keep you up at night shouldn't end up in a discovery request. Ask them somewhere that forgets.