Oneflow vs Sirion: how they compare in 2026
Oneflow is a fit for organizations that want to create and negotiate web agreements from CRM records, then review them against workspace rules. Sirion is a fit for a larger contract program spanning repository intelligence, contract creation and continuing management through its Store, Create and Manage structure. AskSirion links answers to contract material, while Oneflow presents clause findings and missing concepts against a configured playbook. Their data permissions differ substantially. Sirion's current terms permit specified customer data processing to inform machine learning, beyond a simple ban on outside providers training their models. Oneflow expressly prohibits training but removes confidentiality and other undertakings for generative AI functionality. In the difference between vendor learning and provider restrictions, AI Legal Index identifies a contract issue separate from deployment architecture. Oneflow's public entry package also cannot stand in for a Sirion enterprise quote.
At a glance
All 15 axes, side by side
The same grid applied to every vendor in the index, graded from public sources. Hover a grade to see what the letter means on that axis.
AI Centrality
How much of the product is actually AI. Whether the machine learning is the mechanism the buyer is paying for or a feature layered onto conventional software, and whether the vendor is specific about which is which.
AI supports several substantial contract tasks within Oneflow's wider platform. Write with AI drafts, edits and translates text inside an agreement; AI Extract imports participants, dates and duration into editable metadata. AI Review compares terms with a workspace playbook, and AI Insights aggregates the resulting risks across the portfolio. The premium package adds custom rules, dashboard widgets and a report for negotiation follow up. Electronic signatures, editable HTML agreements, permissions, approvals and renewal reminders continue to work without these models. Customers can disable AI, and several review capabilities remain in beta. AI therefore changes how users analyze contracts without replacing the underlying contract system.
Sirion's contract lifecycle platform came before its AI, and the company has traded since 2012. The product is organized as Store, Create and Manage, covering repository, drafting, approval, negotiation, obligation tracking and performance management. Without the agents and agentOS, a working enterprise CLM remains, and Gartner classifies it as one. The models drive the capability Sirion now sells on top of it. Agents are described as extracting and normalizing contracts, assembling first drafts from a playbook, proposing redlines on the other side's paper and monitoring obligations. Customers can also build and deploy their own agents in agentOS.
Citation Accuracy and Hallucination Disclosure
Whether the vendor publishes measured accuracy on citations and assertions, grounds output to primary sources, and says plainly what its system does when it does not know. Legal has a documented public record of fabricated citations reaching filed briefs, so an untested claim of accuracy is not evidence.
Review findings connect to highlighted passages in the agreement, with controls for moving between them. Each finding carries the relevant playbook concept, configured feedback and example clause text. A separate view identifies required clauses that the review did not find, while portfolio charts open the corresponding contract list. This lets the reviewer compare a proposed concern with the actual wording and the organization's rule. Oneflow warns that generated text can misstate facts, numbers and law and recommends human correction. No reproducible accuracy study, error rate or jurisdiction specific evaluation accompanies these features. The contract highlighting supports source verification, not validation of legal authority.
AskSirion answers questions within one agreement or across the enterprise contract corpus, combining relevant clauses with related obligations, service levels, and counterparty records. Each answer links back to its underlying source, and the interface explains how it interpreted the user's request. Follow-up questions preserve context, while summaries, financial calculations, and date calculations draw on the same contract information. This establishes a traceable retrieval workflow for the customer's agreements. An externally reproducible accuracy evaluation, error breakdown, or legal citator is not specified. References to an agreement establish the source of a commercial term, rather than verifying that an external legal authority remains current.
Autonomy and Oversight Model
What the system decides on its own, what a lawyer must approve, and whether the vendor documents where the review point sits. A tool that drafts under review and a tool that files without one are different products and different risks.
Administrators enable AI Review for a workspace and configure its playbook, including which clauses matter, which findings count as risks or issues, and what feedback appears. Role permissions separately control access to individual reviews and the Insights dashboard. Sent contracts receive continuing analysis; draft contracts require a manual review action. A person can inspect highlighted text, edit the playbook and use example wording when revising an agreement. The AI appendix advises human review before relying on output. No confidence cutoff, mandatory lawyer approval after an AI finding, or correction route for an already approved mistake is specified. Risk colors classify findings rather than stop execution.
AskSirion interprets a request and orchestrates agents for drafting, redlining, extraction, obligation tracking, and invoice reconciliation. Custom agent setup uses ordinary language to define the task, accessible data, connected systems, and when to escalate to the user. Agents can be tested with real scenarios and refined after launch. Recommendations include explanations and source links that allow review against the underlying contract. This gives administrators an escalation mechanism beyond a general promise of human control. Default stop conditions, the approval interface for consequential actions, and recovery after an incorrect action remain unspecified. The launch sequence permits an agent to go live before subsequent scenario testing.
Operational and Outcome Evidence
Named, dated evidence that the product works in production at real firms or legal departments. Case studies with figures and identified customers count. Unattributed testimonials and launch announcements do not.
Puzzel's Sales Operations Manager, Anette Johnsen, describes generating offers and contracts from Salesforce opportunities with items from related CPQ quotes. She identifies a concrete benefit of the HTML format: participants revise one agreement instead of uploading successive files before signature. Other named users include Systembolaget's Head of Indirect Sourcing, Tor Myhrman, and Hedin Bil's HR Business Partner, Elin Skoglund, who describe handling agreements and maintaining oversight. These accounts establish operational use across sales, procurement and HR. They do not isolate AI review performance or supply a measured error reduction, evaluation period and sample for Oneflow's current AI packages.
Three customer representatives speak on the record with their roles and employers. Edzard Janssen is Chief Procurement Officer at RBI. Reinhard Plaza-Bartsch is Head of Digital Supply Chain and Operations at Vodafone. Angella Dikmic is Manager of IT Vendor Management at GTAA. All three quotes are qualitative, and all three people work in procurement or vendor management rather than legal. The five percentage claims name no customer and read as product capabilities, so named customers and published figures are never paired. About 25 enterprise logos appear, including Citi, GE, Coca-Cola, Chevron, PayPal, Bayer, Rolls-Royce, Aramco, Yamaha, DP World and Zalando. Customers are in more than 70 countries, with more than $450 billion of contract value under management. Sirion also publishes a case study library.
Privilege and Confidentiality Posture
How client confidences are handled: attorney client privilege and work product treatment, segregation of one client matter from another, whether client data trains any model, and what the vendor commits to in writing rather than in marketing.
The AI appendix expressly prohibits Oneflow from training any AI model on customer data and takes precedence over conflicting general terms for AI features. Ordinary service confidentiality continues while information remains confidential, with restrictions on compelled disclosure. There is a consequential exception: the generative AI section excludes Oneflow's contractual confidentiality obligations along with liability, indemnities and service levels. Review, Insights and Extract have a separate processing section tied to the DPA. Workspace permissions and optional provider controls support access management, but attorney client privilege and work product receive no specific undertaking. The contract therefore requires attention to the exact AI functionality being enabled.
Vendor library articles say customer data stays within the customer environment and is not used to train external language models. Clause 4.2 of the SaaS Terms, version 6, modified 26 August 2026, says otherwise. It reserves a right for Sirion to have its automated systems review and process Customer Data to generally inform machine learning capabilities in the services. None of the published agreements offers an opt out, and the agreement, not the library articles, is what binds. The rest of the confidentiality framework is published and can be read before signing. The data processing addendum stores and processes all customer data in a Sirion application instance specific to the customer. It sets Article 32 technical and organizational measures, with AES-256 at rest and TLS 1.2 in transit. It imposes need to know and least privilege access with automatic logout, requires deletion within 30 days of termination, and commits to notifying the customer of a confirmed security incident within 72 hours. Section 5 of the SaaS Terms adds mutual confidentiality, with a duty to give prior notice of any authority or court demand so the other party can object. The published agreements and trust center do not address privilege or work product.
UPL and Professional Responsibility Posture
Whether the vendor is clear that it supplies a tool rather than legal advice, who its audience is, and how it addresses unauthorized practice of law, competence and supervision duties, and jurisdiction limits. ABA Formal Opinion 512 is the reference point. Where the advice line is not the duty a product raises, the axis is read through the nearest professional duty it does raise: judicial conduct rules and the reviewing duty for products sold only to courts, and the duty to bill for time actually spent for products that draft time entries.
Oneflow's AI appendix says output is not legal advice, legal conclusions, recommendations about legal rights or an application of law to a customer's circumstances. It advises human review because generated content can misstate law as well as facts. A separate UK article maps solicitor responsibilities to SRA standards, confidentiality, client communication and supervision, and describes practical review and pilot controls. North American terms also prohibit treating the platform as a substitute for an attorney. The product nevertheless serves sales, HR and other users without legal qualifications. No comprehensive jurisdiction by jurisdiction set of product restrictions or lawyer competence program is supplied.
The platform is sold to procurement, sales and finance teams as well as in house legal, all working on the same contracts. The three customer referees Sirion features are procurement and vendor management leaders, not lawyers. The home page, the trust center index with its AI ethics, compliance and security sections, and the terms and policies index say nothing on where tooling ends and legal advice begins. There is no statement that Sirion does not give legal advice, no professional responsibility or ethics page, no named bar or ethics guidance, including ABA Formal Opinion 512, and no jurisdiction limits.
AI Governance and Bias Disclosure
Published governance over model behavior: who owns it inside the vendor, what is tested before release, and what is disclosed about disparate output across matter types, parties, or populations.
Oneflow promotes human accountability and warns against automation bias in its guidance for UK legal teams. It recommends evaluating missed issues, overrides and review quality rather than relying only on speed, and says its AI performs best in English and Swedish. These are useful limits and principles for customers designing a pilot. The public material does not identify an accountable AI governance owner, publish a model release evaluation, or report bias and performance differences across languages, contract types or populations. General ISO security controls and a list of security policies concern operational protection; they do not describe a separate AI testing and governance program.
The trust center has a section titled Artificial Intelligence Ethics and Governance, last modified 16 December 2025, and its whole content is three sentences. They say Sirion has set up an AI Governance Program called S-AIGP, which monitors and ensures compliance with the EU AI Act 2024/1689 and other AI regulations as they emerge. Readers are told to contact their Sirion Account Executive for more. Sirion publishes nothing on who owns model behavior, what is tested before release, how the program works, or bias and uneven output.
AI Safety and Data Stewardship
Retention, deletion, access control, and what happens to prompts and documents after they are processed. Whether the vendor states its subprocessors and its incident practice, or leaves the buyer to assume.
Contract data is encrypted with AES 256 at rest and TLS 1.2 in transit. Customers configure permissions and retention policies; administrators can disable optional providers or AI features. Global DPA terms require breach notification without undue delay, while the North American DPA specifies 48 hours. Daily database snapshots last seven days, and system and authentication logs last 90 days. The DPA permits limited residual backups after deletion without fixing that period. Generative AI history can follow provider terms, while the product page promises removal of temporary review and import data. That leaves no single, specific retention schedule for every prompt and output.
The data processing addendum, version 3, modified 2 July 2026, covers retention, access, subprocessors, incidents and encryption. Appendix B sets the retention period by the controller's contractual direction. Clause 8.1 requires deletion or return of all Customer Personal Data, including copies, within 30 days of termination. Any retention the law requires is limited to that purpose and stays under the agreement's protections, and the SaaS Terms add automatic deletion after 30 days as a backstop. Access runs on a need to know and least privilege basis, with automatic logout on inactivity. Subprocessors are named in a list published at its own URL. Sirion gives notice and an objection right for any addition or replacement, imposes equivalent terms on each, and is liable for their breaches. Clause 4.3 commits Sirion to notify the customer within 72 hours of establishing material impact from a confirmed Security Incident, and Appendix B says 48 to 72 hours, with updates until resolved. The definition expressly excludes unsuccessful attempts such as failed logins and denial of service attempts. Encryption is AES-256 at rest and TLS 1.2 in transit.
AI Liability and Recourse
What the vendor stands behind contractually when its output is wrong. Indemnities, caps, carve outs, insurance, and whether any of it is published or only reachable through a negotiated agreement.
Ordinary global service terms include an intellectual property defense obligation, a material performance warranty and a general cap based on 12 months of fees. Those protections do not carry through unchanged to AI. The appendix supplies AI output as is, disclaims warranties and excludes all liability and indemnification undertakings for generative AI functionality, together with confidentiality obligations and agreed service levels. It also permits specified AI suspensions or termination without reimbursing fees. North American terms contain the same generative AI exclusion. A wrong clause, inaccurate summary or reliance on generated advice has no specific affirmative AI remedy in these published terms.
Sirion publishes its SaaS terms and a separate end user agreement, so both can be read before signing. Clause 7.1 commits Sirion to indemnify and defend the customer against third party claims that use of the services infringes intellectual property. Clause 7.2 sets four exclusions, namely combination with materials Sirion did not provide, unapproved changes, unauthorized use, and use inconsistent with the documentation. Clause 7.3 gives the remedies of securing the right to continue, replacing or modifying the services, or terminating with a pro rata refund of prepaid fees, as Sirion's sole and exclusive liability. Clause 7.4 has the customer indemnify Sirion over customer data. Clause 8.1 caps each party's total liability, across the whole term, at the fees paid for the subscription in the twelve months before the first event giving rise to a claim. Clause 8.2 excludes lost profits, lost or damaged data and consequential loss, and under clause 8.3 the cap does not apply to fees owed, breach of confidentiality or the indemnities. Clause 6.2 warrants that the services perform materially in line with the documentation, with a pro rata refund if Sirion cannot correct a nonconformity reported within ten days. Separate Sirion AI Services terms extend the intellectual property indemnity to the AI features, but exclude claims that come from the customer failing to review or validate output. Section 5.1 of those terms disclaims any warranty that output is accurate or free from hallucinations. No insurance position is published.
Practice Systems Integration Depth
How deeply the product reaches into the systems legal work already lives in: document management such as iManage and NetDocuments, Word and Outlook, contract lifecycle management, matter management, e-billing, and court filing systems.
The Salesforce integration creates interactive contracts from opportunities and transfers contact, company and opportunity data into them. Its configurable sync sends data from Salesforce objects into Oneflow and returns contract values to Salesforce. Product tables can reflect price books or CPQ quotes, while workflows map contract stages back to pipeline stages. Engagement events and status changes appear in the Salesforce timeline, and returned contract data supports dashboards. Oneflow identifies separate setup guidance for older and newer integration versions and compatibility with Professional edition upward. HubSpot, Dynamics 365 and other CRM connections complement this workflow; Salesforce and Dynamics can require premium integration charges.
Nine named connectors are documented, each with its direction of travel, in the Non-Native Integrations article in Sirion University, last updated 24 February 2026. Four SAP Ariba accelerators cover procurement, bids, awards and contract workspaces. An executed contract request in Sirion creates a purchase requisition in Ariba, and the resulting purchase order is written back. Ariba sourcing events and RFPs become Contract Draft Requests on set triggers, one per bidding supplier with line items mapped, and supplier redlines move into Sirion for legal review. Awarded RFx events create a draft request with awarded line items synced. Contracts finalized in Sirion copy into Ariba Contract Workspaces with metadata, documents, line items and status. Three more connectors cover supplier master data from Ariba, SAP S/4HANA Business Partners on scheduled transfers, and Oracle ERP with logging, error handling and sync status. The Microsoft Dynamics connector runs both ways in auto or manual mode. The article describes configuration through triggers, modes and scheduling rather than as a full guide, and the wider Sirion University catalog needs a login. Sirion's public pages do not document the Word add in or the Salesforce connector to this depth. No document management connector such as iManage or NetDocuments appears on them.
Deployment Model and Data Residency
Where the software runs and where the data sits. Multi tenant cloud, single tenant, private deployment, on premises, and whether region of residence is a published option or an enterprise conversation.
Production infrastructure uses AWS availability zones in Ireland, with document replication and backups in Sweden. Oneflow's AI FAQ broadly describes data as remaining in the EU, and Insights Premium is described as running within Oneflow's own EU infrastructure. The published subprocessor table gives a more varied picture: optional Legly review operates in Germany, optional OpenAI extraction in the United States, and Postmark email processing also in the United States. Administrators can disable optional services. The public descriptions do not fully reconcile these routes or map every AI function to a processing region. No on premises deployment is offered in the published package descriptions.
Sirion's application is SaaS hosted by a cloud provider, according to the data processing addendum, and all customer data is stored and processed in a Sirion application instance specific to that customer. That isolation commitment sits in a contract rather than in marketing. Four underlying providers appear in the trust center, namely AWS, Azure, Oracle and IBM, which suggests customer choice without offering it as an option. Sirion's published documents make no residency commitment and give no region list or jurisdiction option. Nothing separates where processing happens from where data is stored. A Sirion library article describes the platform's multitenant scalability, which sits awkwardly beside the addendum's customer specific instance, and nothing published reconciles the two.
Security Certifications and Trust Center
Independent attestation a buyer can pull without a sales call: SOC 2, ISO 27001, penetration test summaries, a trust center with current reports and named scope rather than a badge image.
A public certificate issued by Qvalify covers Oneflow AB's digital solution for contract automation under ISO 9001:2015, ISO 14001:2015 and ISO 27001:2022. It was issued on June 28, 2024 and remains valid through June 30, 2027, identifying the Stockholm organization and its registration number. A June 2026 statement of applicability lists the security controls and their implementation status. The security center separately identifies SOC 1 Type 1 and SOC 2 Type 1 reports issued by Grant Thornton, available under NDA. Those reports concern control design; neither is identified as a Type II examination of operating effectiveness over a stated period.
Appendix B of the data processing addendum names ISO 27001 as a held certification and describes the review process around it. Under clause 7.1, unless agreed otherwise, Sirion will give a copy of its latest security attestation report on the customer's written request, no more than once a year. Clause 7.2 adds a customer audit right, with scope, timing and controls agreed in advance and a reasonable fee. Appendix B also says Sirion reviews its cloud providers' SOC 1, SOC 2 and ISO 27001 reports as part of a shared responsibility model. That is oversight of the infrastructure rather than an attestation over the application. The report route is open to existing customers, so a prospective customer cannot get the report before signing. The addendum names no auditor, coverage period or scope statement for the ISO certificate. Sirion's trust center states ISO 27001:2022 certification. Its SOC 1 Type II and SOC 2 Type II sections explain what those reports cover, and its security page attributes SOC 1 and SOC 2 to the cloud providers Sirion runs on.
Model Supply Chain Disclosure
Which models sit underneath, whose they are, where they run, and whether the vendor commits to telling customers when that changes. A legal buyer inherits every dependency it cannot see.
The subprocessor list identifies Legly AB for contract review in Germany and OpenAI for extraction in the United States, alongside AWS infrastructure in Ireland and Sweden. Both AI suppliers are optional and controlled through account settings. The AI appendix directs customers to their Data Management page for the current generative model suppliers and their terms. It does not publish model versions or a complete function to model map. DPA provisions require reasonable advance notice of subprocessor changes to the customer's designated email address and allow a 14 day objection. That process concerns suppliers, without promising notice of every underlying model change.
A combination of more than ten large language models and more than 1,200 proprietary small language models powers the contract agents. A routing framework selects a model according to the task, while company content, policies, and risk preferences can refine agent behavior. The same architecture supports extraction, drafting, redlines, obligations, and invoice reconciliation through agentOS. Provider names, model families and versions, and processing locations for individual inference tasks are not identified. There is also no model change notification commitment attached to this architecture description. The disclosure explains how different model types are combined without identifying every external model a customer inherits.
Commercial Transparency
Whether a buyer can learn what this costs without entering a sales process: published rates, the unit being charged, what sits behind an enterprise tier, and what implementation adds.
Business starts at $292 a month, billed annually, including five users; Enterprise has ten users and a custom quote. Business includes writing, extraction and summary features. Standard AI Review and Insights have conflicting public treatment: the pricing page marks paid beta additions, while September help guidance says they are included in Enterprise during beta. Insights Premium is explicitly a paid addition for Business and Enterprise. SMS, identity verification, premium integrations and enhanced support can cost more. A 14 day trial needs no card. Implementation has no published amount, and global annual renewals can increase by the greater of 5 percent or Swedish CPI.
No pricing is published at any level, including the unit of charge. Sirion's site has no pricing page, no tier structure, no unit per seat, per contract or per agent, no volume banding and no statement of what implementation adds. Every call to action is a demo request. No published figure lets a prospective customer estimate cost before a sales process, even though Sirion publishes its SaaS terms, end user agreement and data processing addendum in full.
Firm and Practice Coverage
Who the product is actually built for. AmLaw, midlaw, small firm and solo, in house departments, government and courts, and which practice areas are supported rather than merely claimed.
Legal, procurement, sales, finance and HR teams can create and manage agreements in Oneflow. Legal users configure review rules and approve terms; business teams use reusable templates and CRM data for offers, sales agreements and employment documents. Named customers span telecoms, retail, recruitment and public sector purchasing. UK guidance also addresses solicitors using AI for commercial contract review, bringing professional duties into the intended audience. AI is optimized for English and Swedish, with variable results in other languages. The product does not supply a jurisdiction specific research corpus, and its coverage statements do not define supported litigation work or every category of legal agreement.
Eight industries have their own pages. Financial services is split into procurement and into capital markets and lending. Insurance has a separate solution for triaging underwriting submissions. The others are automotive, IT services, healthcare, pharmaceuticals and life sciences, telecom, and oil and gas. Five departments have their own pages, covering in house legal, legal operations, procurement, sales and finance. Customers are in more than 70 countries, and the site is in English, German and French. Sirion does not say where the product stops. It excludes no contract or matter types, does not address law firms, government or public sector use, and describes coverage by industry and department rather than area of law.
The 12 legal signals, side by side
Recorded rather than graded. These are the questions a practitioner has to answer before a tool touches a client matter, and the answers are taken from public material only.
Client Data in Training
Can material a lawyer puts into this product be used to train a model?
The AI appendix forbids Oneflow from using customer data to train any AI models and overrides conflicting general terms for AI issues. Inputs and attributable outputs fall within the customer data definition. The general service license permits improvement and anonymized analysis, and AI feedback may improve accuracy, but those clauses do not remove the express training ban. The product FAQ also says underlying models do not train on customer data.
Separate provider processing, retained history and the generative AI confidentiality exclusion remain relevant to handling client material.
Library material from Sirion states that customer data stays within the customer environment and is not used to train external language models. Its SaaS Terms, version 6, modified 26 August 2026, reserve the opposite at clause 4.2. Sirion may direct its automated systems to review and process Customer Data to generally inform machine learning capabilities in the Subscription Services. The same clause covers generating aggregated and anonymized industry analytics, with a commitment not to publicly identify the customer or disclose Customer Data to third parties.
The agreement governs over a policy page, so the reservation is what binds the customer. It operates on Customer Data itself rather than on deidentified or aggregated derivatives, which gives it a broader reach than clauses limited to aggregate data. It says generally inform rather than train, so the words leave open whether it authorizes model training in the ordinary sense. The SaaS Terms, the end user agreement and the data processing addendum offer no opt out, consent step, configuration setting or exclusion route.
Prompt and Output Retention
How long does the product keep what a lawyer typed, and can that be set to zero?
Oneflow permits customers to set retention policies for stored contracts, and its AI page says temporary import, review and Insights data is removed after use. Generative AI has a different contractual provision: input questions and output responses may remain as history under the model supplier's current terms. No uniform period covers that history or every provider copy. Daily database snapshots last seven days, system logs last 90 days, and residual backups after termination have no precise DPA deadline. Contract deletion therefore cannot be treated as an immediate purge of all AI processing records.
The published data processing addendum says the retention period is set by contractual obligations as the controller directs. After termination, personal data is typically kept for 30 days. The window is therefore under customer instruction, with a stated default, in a contract rather than a policy page. The provision covers personal data rather than prompts and generated outputs specifically, and no zero retention setting is offered.
Ethical Walls and Matter Segregation
Does retrieval respect the firm’s ethical walls, or can the model read across them?
Workspace roles control access to contracts, and Oneflow provides distinct permissions for AI Review and the Insights dashboard. Administrators enable review for selected workspaces and can disable AI features or optional suppliers. These controls can support departmental separation and limit who sees review results. The descriptions do not specify how retrieval enforces every contract permission within a workspace, how restrictions propagate to generated summaries, or how a law firm's matter wall follows an AI request. Permission controls alone do not establish a complete ethical wall mechanism for confidential client matters.
Separation between customers is set out in the data processing addendum rather than in marketing. All customer data is stated to be stored and processed in a Sirion application instance specific to the customer, within the cloud service. A shared responsibility model applies, in which Sirion secures the software, the customer data and the related access, and the cloud provider secures the underlying facility. Vendor library articles describe role based access control with granular permissions.
Sirion's published material does not describe how access is enforced between teams inside a customer, where legal, procurement, sales and finance share the platform. A Sirion library article also describes the platform's multitenant scalability, and nothing published reconciles that with the addendum's customer specific instance. The buyer is an in house department, so separation between customer tenants is what applies here.
Third Party Request and Subpoena Notice
If someone subpoenas the vendor for a firm’s data, does the firm hear about it first?
Global confidentiality terms require reasonable prior notice when disclosure is legally compelled, unless notice is prohibited, and reasonable assistance at the customer's expense in seeking protection. That gives the customer an opportunity to contest or narrow a request involving ordinary confidential service data. The AI appendix expressly removes contractual confidentiality obligations for generative AI functionality, so the same protection should not be assumed across every AI route.
No transparency report accompanies the commitment. North American and global agreements need to be matched to the contracting entity and enabled services.
Section 5.3 of the SaaS Terms, version 6, modified 26 August 2026, holds the notice commitment. A party receiving a demand from an authority or court for the other party's Confidential Information may comply only after satisfying itself the demand is lawful. It must give as much prior notice as possible, where possible, so the other party can object, and mark the material as the other party's Confidential Information.
The duty is mutual and covers customer material. Confidential Information includes information that should reasonably be understood as confidential, and the AI Module clause at 1.8(b) confirms this by bringing Customer Data back into that definition. The notice duty is limited by the words where possible, not by a legal prohibition. The data processing addendum does not address third party demands. The terms pages, the addendum and the trust center give no transparency report or count of demands.
Primary Law Corpus Provenance
Where does the law in this product come from, and does the vendor have the right to use it?
Oneflow's review and portfolio analysis work with the customer's agreements and configured playbook concepts. Example clauses come from the organization's rule setup, and drafting can use selected text as input. No catalog of legislation, reported decisions or licensed legal commentary is supplied for these features. Naming a model provider does not identify the legal sources in its training data or their licensing terms.
The UK educational articles link professional guidance, but those links do not constitute a legal research corpus inside the product.
The working corpus is the customer's own contract set. Answers are described as carrying citations linked to the customer's own data, and agents extract and normalize contracts from sources the customer nominates. Sirion separately states that its purpose built models are trained on millions of enterprise contracts. That describes a second corpus by its makeup, but Sirion names no source, states no license or rights basis, and does not say whose contracts those are.
The product does not retrieve primary law, so the usual questions of jurisdiction and coverage do not arise. No update cadence is published.
Good Law Verification
Does the product tell you when the authority it just cited has been overruled?
Highlighted review passages let a user inspect contract language, not whether a judicial decision remains authoritative. Oneflow has no published citator connection, treatment history service or automated check of legislation's current status. Teams maintain their own playbooks, including preferred and fallback clauses, so changing a rule after a legal development remains an organizational responsibility. The appendix warns that AI can generate incorrect law and recommends human review. A portfolio risk chart or missing clause flag does not establish that the underlying legal proposition remains valid.
The platform manages contracts, obligations and supplier performance rather than retrieving case law or legislation, so a citator is not part of what it sells. The home page, the trust center index with its AI ethics, compliance and security sections, and the terms and policies index do not address whether legal authority is reviewed for subsequent history. They describe no citator, treatment signal or currency check.
Refusal and Uncertainty Behavior
What does the product do when the answer is not in the corpus?
The AI Review interface separates findings from a missing clauses view. That view identifies playbook concepts absent from the agreement, giving the reviewer an explicit negative result instead of merely omitting a match. Red and yellow labels distinguish issues from risks; they are not numerical confidence estimates. Oneflow warns that generated facts and law can be wrong, but does not publish a general abstention rule for drafting or summaries. Missing clause handling is described for review, without promising that every unsupported question will be refused.
AskSirion gives source references and explains how it interpreted a request when answering from contracts and related records. Custom agents can also escalate according to conditions the customer specifies. Neither feature establishes a response for a question unsupported by the corpus, or an uncertainty indicator presented to the reviewer. Contract issue severity and an administrator's escalation policy concern different decisions from whether an answer has adequate grounding.
Fabricated Citation Record
Does a public court record exist addressing fabricated or hallucinated legal citations in output from this product?
Damien Charlotin's AI Hallucination Cases database contained no entry naming Oneflow as of October 16, 2026. The result covers the database's reported court incidents, not every private use of the product or every inaccurate contract response. Oneflow itself discusses court examples involving generic AI tools in educational material; those examples do not identify Oneflow as the tool involved. Contract drafting and risk analysis can still produce incorrect law, which the AI appendix expressly warns about. The absence of a named incident provides no accuracy guarantee.
Sirion and its former name SirionLabs have no matching entry in Damien Charlotin's AI Hallucination Cases database. The database records judicial decisions addressing fabricated AI content and identifies an implicated tool when the case supplies one. Sirion's enterprise contracting and supplier governance functions differ from a litigation drafting service. The absence of a named database entry does not establish that every extracted term, contract answer, or proposed revision is accurate.
Bar Guidance Alignment
Has the vendor engaged in public with the ethics opinions its buyers are bound by?
Oneflow's UK contract review guidance maps SRA standards and conduct duties to human supervision, confidentiality, client transparency and practical controls. It also discusses insurance, data protection and the legal team's responsibility for final judgment. This is a jurisdiction specific mapping with suggested review steps, rather than a generic statement that lawyers remain responsible. It is vendor guidance, not regulator approval or an independent finding of compliance. The mapping does not establish equivalent coverage for every jurisdiction in which Oneflow sells its platform.
Sirion's compliance material is oriented to regulation and security frameworks, covering ISO 27001, SOC 1, SOC 2 and the EU AI Act. None of it addresses the professional conduct obligations binding the lawyers who use the product. The home page, the trust center index with its AI ethics, compliance and security sections, and the terms and policies index do not engage with any bar or ethics guidance. That includes ABA Formal Opinion 512 and any state bar or Law Society material.
Billing and Fee Posture
Does the vendor address what happens to the bill when the work takes an hour instead of six?
The vendor's UK AI review material addresses solicitors as well as internal legal departments and promotes faster review with retained professional judgment. Its product pages also describe reducing routine contract work and administrative effort. No accompanying procedure explains how a firm records AI time, adjusts a client invoice, discloses savings or passes through Oneflow charges. Subscription pricing is the buyer's software cost rather than a client billing policy.
Because the audience includes lawyers serving clients, the time savings claims cannot be treated solely as an organization's internal operating expense.
Public materials are built around speed and automation gains. They cite 90 percent faster contract centralization, 85 percent faster insights, up to 90 percent faster time to contract and 70 percent faster agentic automation. A customer reports as much as a 50 percent rise in tasks automated. The home page, the trust center and the terms and policies index offer no record of AI assisted work by matter for fee purposes, and no guidance on billing, fee or disclosure treatment.
The buyer is an in house function rather than a firm billing a client, so fee disclosure here comes down to internal cost and outside counsel spend, and the published material addresses neither.
Outside Counsel Guideline Readiness
Can a firm get this vendor through a client’s AI clause without a bespoke negotiation?
Oneflow supplies a public DPA, global and North American service terms, an AI appendix and a subprocessor table with functions and processing locations. Optional AI suppliers include Legly and OpenAI, and administrators can control optional services. A public ISO certificate and statement of applicability support security diligence; SOC reports require an NDA. These materials help counsel explain data handling to a client.
They do not form a dedicated client AI disclosure package, and the generative AI confidentiality exclusion and conflicting regional descriptions require specific attention in any outside counsel approval process.
Appendix D publishes a Sub-Processor List, stated to be regularly updated, at a named URL that can be reached without a sales conversation. Clause 5.4 commits Sirion to notify the customer of updates and give an opportunity to object to additions or replacements. Clause 5.6 requires equivalent data protection terms on every subprocessor and keeps Sirion liable for their breaches. Sirion publishes no disclosure pack for a firm answering a client's AI clause, and no consent or notification material for clients.
Its published material names no model provider. The trust center's AI ethics page routes AI governance detail to a Sirion Account Executive.
Court Disclosure Support
If a judge’s standing order requires an AI disclosure, can the product produce one?
AI Insights supports an export of portfolio analysis, and reviewers can inspect the contract passages associated with a playbook finding. Insights Premium adds an action report for internal discussion and negotiation. Enterprise also includes an account audit log, while signing workflows maintain transaction records. The published export instructions do not establish a complete record of model identity, prompt, output and human edits for each AI action.
These are useful contract administration records, but no court disclosure template or complete model provenance export is described for a later filing.
Answers are said to carry citations linked to the customer's own data. Vendor material says every AI response includes citations to exact sources, with citation and justification at clause level. Audit trails are said to record system activity, user actions and data changes. Together these cover the sources retrieved and part of what was done, as a side effect of the product's design. Sirion identifies no model, so its material does not show which system produced a passage, and it describes no export built for court disclosure or AI use certification. The product serves enterprise contracting rather than litigation, so court standing orders are not its usual setting.
The questions both sides leave open
Derived from the records above rather than written, so it cannot favor either vendor. Take these into both conversations and ask each side the same question.
- Good Law Verification
Which one fits
Choose Oneflow if
- Shared web negotiation and Salesforce synchronization define the project.
- An explicit AI model training prohibition is required.
- The team wants a published Business starting package for five users.
Choose Sirion if
- Repository intelligence, creation and continuing management form one enterprise program.
- AskSirion's linked contract answers fit the legal review process.
- Customer application isolation is part of the desired deployment design.
In summary
Oneflow
Oneflow connects CRM generation, live editing and signing with playbook review and portfolio charts. Its AI terms prohibit model training on customer data while imposing separate generative AI exclusions.
Sirion
Sirion organizes contract work through Store, Create and Manage with agent capabilities and cited answers. Its deployment descriptions include customer application isolation and several enterprise cloud relationships.
Questions buyers ask
Does Sirion's outside provider training restriction cover its own learning?
No. Its current terms separately permit specified processing of customer data to inform machine learning. Restrictions on external model providers should not be extended to erase that own service permission. From the AI Legal Index, based on each vendor's own published materials as of October 16, 2026. No vendor pays for placement.
Are Oneflow's risk charts the same as AskSirion?
Oneflow charts risks by configured playbook concept and opens the corresponding agreement list. AskSirion supports questions with links to contract material, so the reviewer reaches an answer through a different interaction. From the AI Legal Index, based on each vendor's own published materials as of October 16, 2026. No vendor pays for placement.
Can Oneflow guarantee EU processing across every optional service?
Its core storage is in Ireland and Sweden, but the published supplier list includes optional OpenAI extraction and Postmark processing in the United States. Enabled services matter to the actual processing route. From the AI Legal Index, based on each vendor's own published materials as of October 16, 2026. No vendor pays for placement.
Which contract distinction should lead the data discussion?
When comparing the scope of model learning rights, AI Legal Index separates Sirion's own machine learning permission from limits on external providers. Oneflow prohibits training but has a separate generative AI confidentiality exclusion that also needs an explicit procurement decision.
Neither vendor paid for inclusion, placement or a grade, and neither reviewed this page before it published. Everything on it comes from public material on the dates shown. How the index grades.
Sirion's customer isolation describes deployment architecture, not an automatic prohibition on permitted data learning.