Contracts & Legal

Find critical clauses before they get expensive.

Viktor reads your contract portfolio and checks every clause against current law. Every statement names contract and section.

The problem

Hundreds of contracts meet new law – and time is short.

Contracts sit in folders, amendments in emails, and the law keeps changing. No search shows you which clause is now a problem.

The question

You ask Viktor – in one sentence.

Instead of building reports, you simply ask. Viktor shows what gets checked – step by step, across all sources.

The evidenced answer

Every finding comes with its citation.

Every statement names the source it comes from. Whatever can't be evidenced, the agent marks as open.

The next step

Viktor prepares – you approve.

Viktor prepares the next step. It is only carried out once you approve it.

Hi, I'm Viktor – your AI agent for contracts & legal. I keep an eye on your systems and speak up when something stands out.
I noticed four signals today – each inconspicuous on its own:
Signals · today
SharePoint148 contracts
OutlookAmendments by email
Jira7 open legal tickets
c:node GraphEU Data Act
@Viktor, which of our contracts do we need to adapt because of the EU Data Act?
I'm checking all sources:
  • Read 148 contracts
  • Filtered 23 cloud, maintenance and data contracts
  • Checked clauses against the EU Data Act
  • Checked data protection annexes (GDPR)
You need to adapt 4 of 23 cloud, maintenance and data contracts. Most urgent:
  • Cloud contract “Hosting Nord” blocks switching providers for 24 months
  • Maintenance contract: machine data belongs to the manufacturer only
  • Two service providers without a data processing agreement
Framework agreement Hosting Nord § 14EU Data Act Art. 25GDPR Art. 28
I've prepared the next step:
Draft · Amendment Hosting Nord
Draft amendment to § 14: switching period of no more than 2 months, data export at no extra cost – for review by the legal department.
ApproveEditNothing goes out without approval

Example case with sample data.

c:node Research · Contracts & Documents

Why critical clauses get overlooked — and how traceable AI binds every statement to its citation.

Reading time ~5 minBasis 6 sources, verifiedTopic Contract analysis & clausesApproach source-backed · on-prem possible
A contract being signed
Fig. 0 · Liability, termination, SLA, price — the decisive clauses are spread across contract and annexes, often in the small print.
Summary

The expensive clauses — liability, notice periods, SLA, price adjustment — are spread across contract, annexes and framework agreements. Whoever has to review quickly misses one. Full-text search finds words, but doesn't connect them. General AI summarises, but invents details. This article shows, with sources, why that is risky — and how c:node checks your contract portfolio against current law and binds every statement to its citation in the contract.

1 The problem: the contract is known — the chain isn't.

A framework agreement refers to annexes, the termination clause depends on the term, liability is limited by the SLA, the price by an adjustment clause. The individual sentences are readable — what makes them risky is their chaining across documents: which clause overrides which, which deadline follows from which term? Exactly these connections are missed during manual review, especially under time pressure. The danger isn't the single sentence, but the overlooked connection.

2 Why full-text search and generic AI don't solve this reliably.

A full-text search finds the passage “liability”, but it doesn't tell you that §8 is limited by the SLA in §6 — multi-step relationships across documents are its weakness.1 An AI chatbot summarises fluently, but the US standards institute NIST states that fabricated statements are inherent to the design of such models — it adds clauses that aren't in the contract, complete with convincing wording.2 In a contract, an invented clause is a liability risk. Even with access to the real text, simple models still add fabrications.5

3 The approach: bind every clause to its citation.

c:node reads in contract, annexes and DMS and lets c:node reason over a knowledge graph — clauses, deadlines, terms and rules as a network of nodes and edges. Research calls this combination of pattern recognition and firmly checked rules neuro-symbolic AI; it delivers answers that can be justified and traced.3 Via the graph, c:node follows multi-step clause relationships — liability §8 → limited by SLA §6 → applies for a 24-month term — which plain text search cannot do.1 Every statement carries its section with it.

:node Graph— Contract analysislive
Source (API)ClauseSummaryRule / citation
Nodes active · edges fire in real timeisolated per tenant · grows with every contract
Fig. 1 · Clauses, term, deadlines and rules connect into a traceable summary. c:node fires the path — every statement stays bound to its section.

4 In practice: scattered clauses become a traceable summary.

The same question that today costs hours of cross-reading, c:node answers in seconds — and discloses every statement:

What are the critical clauses in this contract?

Three points. Termination (§12): only at the end of the term (24 months), otherwise automatic renewal. Liability (§8): limited by the SLA in §6 — only applies at <99 % availability. Price (§4): annual adjustment to an index. Recommendation: set the termination window in the calendar.
DMSContract §12 · §8 · §6 · §4AnnexesClause rule

5 Why this also holds up legally and organisationally.

Because c:node backs every statement with section and history, every summary can be audited afterwards — against the original text, without you having to disclose the inner workings of the model.6 Anything that isn't in the contract is marked as open instead of asserted. And because c:node runs on-prem or in the EU cloud, sensitive contract data stays in the EU or in your own data centre. Data protection is built in, not bolted on.

In short: The risk isn't the single sentence, but the overlooked connection — and the invented clause. c:node binds every statement to its citation. Whatever can't be evidenced, it marks as open — so everything stays auditable.

Sources

  1. GraphRAG-Bench — graph-based AI outperforms plain text search in multi-step reasoning across documents. arXiv:2506.02404, 2025. arxiv.org…
  2. NIST, Artificial Intelligence Risk Management Framework: Generative AI Profile (NIST AI 600-1), 2024 — fabricated statements are inherent to the design. nvlpubs.nist.gov…
  3. Hitzler et al. (eds.), Handbook on Neuro-symbolic AI and Knowledge Graphs, IOS Press, 2025. iospress.nl…
  4. MEGA-RAG (Multi-Evidence RAG) — retrieval reduces fabricated statements by more than 40 %, but does not eliminate them. PMC12540348, 2025. ncbi.nlm.nih.gov…
  5. Retrieval-augmented generation significantly reduces fabricated statements, but does not eliminate them entirely. arXiv:2404.08189, 2024. arxiv.org…
  6. European Union — EU AI Act, High-level Summary: logging, documentation, human oversight for sensitive AI. artificialintelligenceact.eu…

The sources support the principles (multi-hop across documents, limits of generative AI, neuro-symbolic approach, retrieval reduces errors, EU AI Act requirements). Clauses and sections in the live graph and the sample dialogue are simplified illustrations, not real contracts.

Get started

Three ways to c:node.

Same product, same evidence – you choose where it runs.

Check your own contract — traceably.

Give c:node a small test run with a real contract, or talk to the team behind it.