CHLLM is in beta. Tell us what is missing: write to us →
01 · A REPORT IN DETAIL

The AI that investigates — proof in hand.

You ask a question. CHLLM refines it with you, investigates across hundreds of sources, then delivers within a few dozen minutes a report where every claim opens onto its source. Nothing to take on faith: everything can be checked.

for lawyers, analysts, consultants, researchers · EU-hosted
Excerpt from a real report (in French) — read in full →
92 sources 7.9k words 30m43s REPORT

Personal liability of a French SAS director in judicial liquidation

The action for insufficiency of assets requires a management fault that contributed to the shortfall [art. L. 651-2 C. com.]. Mere negligence by the director has been excluded since the Sapin II law, on a broad reading favourable to directors acting in good faith [Cass. com., 3 Feb. 2021]. The catalogue of faults remains well stocked: missing accounts, carrying on a loss-making business — even when committed during a prior receivership [Cass. com., 22 Jan. 2020].

▸ hover a citation to see its source▸ every citation links to its source, automatic demo
02 · THE DIFFERENCE

What a chatbot cannot prove to you

A chatbot answers from memory, in seconds: plausible, but unverifiable. CHLLM runs the investigation at the moment you ask — and puts the sources in front of you.

An AI chatbot
Answers from memory, without reading anything
Sources missing, sometimes invented
One single pass, in seconds
A plausible answer, take it or leave it
CHLLM
Searches and reads the sources at the moment you ask
Every claim cited — the source opens in one click
Agents in parallel, hundreds of pages read
A structured report you can check point by point

This search, step by step, replayed

~31 min of real work, replayed in 12 s — reconstruction Read the full report (in French) →
USER QUESTIONLiability of a SAS director in judicial liquidation
SQ 1 · ASKED BY CHLLMWhich texts ground the insufficiency-of-assets action?waiting…
SQ 2 · ASKED BY CHLLMWhich management faults are retained in practice?waiting…
SQ 3 · ASKED BY CHLLMIs mere negligence enough since Sapin II?waiting…
CROSS-CHECKConfirm the appeal numbers of 3 key rulingswaiting…
REPORT — IN PROGRESS 0 sources
I. Applicable texts (art. L. 651-1 to L. 651-3)

[L. 651-2 C. com.][legifrance]
II. Management faults retained in practice

[Cass. com., 22 Jan. 2020][+9 rulings]
III. Mere negligence is exonerating (Sapin II)

[Cass. com., 3 Feb. 2021]
IV. De facto directors, corporate presidents

[Cass. com., 9 June 2022][art. L. 227-7]
V. Liability to third parties: the separable fault

[Cass. com., 20 May 2003][final cross-check]
0 OPEN QUESTION — REPORT COMPLETE · 92 SOURCES
REPORT — IN PROGRESS 0 sources
I. Applicable texts (art. L. 651-1 to L. 651-3)
II. Management faults retained in practice
III. Mere negligence is exonerating (Sapin II)
IV. De facto directors, corporate presidents
V. Liability to third parties: the separable fault
0 OPEN QUESTION — REPORT COMPLETE · 92 SOURCES
0 agents at work: an Orchestrator, Investigators, Editors
0 source pages read per report on average
0 sources cited on average
0 specialised modes: legal, academic, scientific, medical, general
averages observed in deep mode
03 · THE SOURCES

You choose where CHLLM searches

A mode focuses the investigation on the sources that carry authority in a field. Five modes cover the main grounds. You can also build your own.

Lawlégifrance · cassation · conseil d'état
Academiccairn · hal · persée
Scientificpubmed · arxiv · nature
Medicalhas · who · cochrane
Generalthe whole web
+ Build your ownthe domains you pick

List the domains you trust. CHLLM favours that scope. Any citation outside it is flagged in the report.

04 · THE METHOD

How a search unfolds

01 · ORCHESTRATOR You ask your question

The Orchestrator reads it and refines it with you if needed — two or three questions, not an interrogation. It recommends a depth, you confirm, the investigation starts.

02 · INVESTIGATORS The Investigators set off in parallel

Each digs into its own angle, all at the same time. They cross-check what holds, note where every piece of information comes from, and unearth the questions you had not thought to ask.

03 · EDITORS The Editors assemble the report

Outline, clickable citations, list of sources. Meanwhile, you are elsewhere: the research runs on our servers and the report waits for you. Allow ten minutes to an hour.

The full guide →

05 · IN PLAIN TERMS

The rules of the game

Three levels of analysis

Éco dispatches the simple questions. Standard covers most searches. Profond calls on the largest models, for subtle or highly technical subjects. The exact cost is displayed before you launch, and a failed search is not charged.

Everything can be checked

An AI tool can be wrong. So can CHLLM. The difference: every claim links to its source — you open, you read, you judge. And any citation outside your trusted perimeter is flagged with a ⚠.

Private, hosted in Europe

Your searches stay private, hosted in the European Union. They only become public if you create a share link — which you can revoke at any time.

06 · THE TRIAL

Judge on the evidence: 50 free credits

50 credits at sign-up — €70 of documentary research, no credit card. Run real searches on your real cases, then judge on the sources.

0free credits
=€70of documentary research, free

Granted when you create your account. Test CHLLM on your real files, then decide.

Valid once, for any first registration.
NO CREDIT
CARD
Create my account
07 · FREQUENTLY ASKED QUESTIONS

Frequently asked questions

How is this different from ChatGPT or Perplexity?

A chatbot answers in seconds, from memory or after skimming the web. CHLLM takes tens of minutes: it refines your question, launches several agents in parallel and reads hundreds of real pages. The result is a report where every statement is cited. You can verify what you read.

Are the reports reliable?

Every statement links to its source, which you can open and verify. Like any AI tool, CHLLM can be wrong. That is why everything is cited. Cross-check critical information before any decision.

How long does a search take?

From ten minutes for a quick answer to about an hour for an in-depth study, depending on the level you choose. You can close the tab: the search continues on our servers and the report waits for your return. A notification alerts you if you have enabled them.

How much does it cost?

CHLLM runs on credits. A search consumes credits according to its depth, and the cost is displayed before every launch. Trial credits are granted at sign-up. After that, you top up with packs, no subscription. A failed search is not charged.

Where does CHLLM search?

Wherever you decide. Modes focus the search on the sources that carry authority in their field: Légifrance and the Cour de cassation for French law, PubMed for medical questions. The general mode opens the whole web. You can also build your own list of domains. Any citation outside your scope is flagged in the report.

Can I steer the search before launching it?

Yes. The framing Assistant reads your question and asks two or three questions if it is ambiguous. It rewrites a precise brief and recommends a level. You confirm or adjust before launch. Framing is free.

Are my searches confidential?

Your searches are private and hosted in the European Union. They only become public if you create a share link. You can revoke it at any time.

What language are the reports in?

French or English, your choice on every search. Source excerpts stay in their original language: an English report can quote a French ruling in the text.

Does CHLLM replace a lawyer, a doctor, a consultant?

No. CHLLM is a documentary research tool. It gathers and cites sources. It gives no legal advice, no medical opinion, no investment recommendation. Professional judgement stays yours.