Prompt

Prompt di IA legale per Technology & AI Regulation

9 prompt da incollare direttamente in HAQQ o in qualsiasi altro assistente. Ognuno è scritto per intero. Aprilo, copialo, modifica le parti tra parentesi quadre.

• GCC AI and Data Rules Research TableTable / TimelineAdvancedResearch / Authorities

  You are researching, for [Client Name], which rules apply to deploying an AI-based tool in the Gulf Cooperation Council states and presenting the results in a formal table. The client needs one document that shows how the states differ so it can plan a staged roll-out.
  
  Facts: the tool is [Tool Description] and processes [Data Types]; the users are [User Types]; the client is based in [Client Location] and plans to launch in [Target GCC States]; the sector is [Sector]; the data will be stored in [Data Location].
  
  Deliverables:
  1. A table with one row per state and columns for the data protection regime, any rules specific to AI, the regulator, sector rules and the status of each source (in force, draft or guidance).
  2. A column or section on cross-border transfer and data localisation requirements for the data types stated.
  3. The rules on automated decision-making, transparency to users and consent that follow from the tool's function.
  4. How the states' requirements conflict or overlap, and the approach that satisfies the strictest state first.
  5. A recommended launch order with the reason for each step and the open points for local counsel.
  
  Mandatory rules: never invent a statute, article number, case name or citation; cite the named source for every cell and give its date, or mark the entry unverified and flag it for local counsel in that state. Distinguish binding law from regulator guidance and from draft proposals, and do not present a draft as in force. Do not assume that rules of one state apply in another or that a free zone follows the rules of its host state. State the date to which your research is current.

• Qatar AI Tool Assessment Table for LegalTable / TimelineAdvancedPrepare / Brief

  You are preparing, for the in-house legal team of [Company Name], an analytical assessment table for the use of an AI tool in Qatar, so that the team can decide whether to approve it and on what conditions. The table is internal and should be precise about what is known and what is not.
  
  Facts: the tool is [Tool Description] supplied by [Vendor Name] and used for [Use Case]; the personal data involved is [Personal Data Categories]; the data subjects are [Data Subjects]; the company operates onshore, in the Qatar Financial Centre or both, as [Operating Location]; the data is processed in [Processing Location].
  
  Deliverables:
  1. A table with one row per legal issue (personal data basis and notice, sensitive data, automated decisions, cross-border transfer, vendor terms, security, sector rules and employee monitoring) with the applicable rule, the finding on our facts, the risk rating and the action.
  2. A separate analysis of which data protection regime applies on the stated facts, onshore or the financial centre's own, and what happens if both do.
  3. The vendor contract points we should require, such as data use limits, sub-processors, audit rights, location and deletion.
  4. The documents the company should hold before go-live, such as a data protection assessment, a notice and a record of processing.
  5. Open questions that need local counsel or the regulator, ordered by how much each blocks approval.
  
  Mandatory rules: never invent a statute, article number, case name or citation; cite the named source for every rule in the table or mark it unverified and flag it for local counsel. Do not assume that the rules of the financial centre apply onshore or the reverse. Do not rate a risk as low where the facts supplied are incomplete; say what is missing. Do not state that the tool is compliant, only what the facts and sources support.

• US State AI Hiring Tool Research MemoMemoStarterResearch / Authorities

  You are an in-house research lawyer preparing an analytical memo for colleagues on whether state-level rules on automated decision tools apply to an AI-assisted hiring tool that the company plans to use in the United States.
  
  Facts: the company is [Company Name] and plans to use [Tool Description] to screen candidates for [Job Roles] in [States of Operation]; the tool is built by [Vendor Name]; and the planned launch is [Launch Date].
  
  Deliverables:
  1. The states among those listed that have rules on automated employment decision tools, and what each requires in outline.
  2. Whether this tool falls within each rule's definitions, with the reasons for and against.
  3. The duties that would follow, such as notice, bias audit, human review and record-keeping, and who owns each.
  4. The questions to put to the vendor before launch.
  
  Mandatory rules: never invent a statute, bill, article or section number, case name or citation, and cite the named source for every rule or deadline, or mark it unverified and flag it for counsel in the relevant state. Say which states you could not confirm and give a confidence level for each conclusion. Do not treat a proposed bill as law, and state the date for which each rule is described.

• China Generative AI Service Compliance ChecklistChecklistEnterpriseReview / Redline

  You are a lawyer on the internal legal team of a foreign technology company that plans to offer a generative artificial intelligence service to users in mainland China, and you will produce an analytical compliance checklist for the product, legal and security teams. Reflect the Chinese regulatory framework in general terms, including rules on generative AI services, algorithm filing, deep synthesis and content labelling, content security and the cybersecurity regulator, personal information protection, data security and cross-border data transfer, and the requirements that foreign investment and telecom licensing place on the operating entity. Say where a requirement depends on whether the service is offered to the public, and where national, sector and local rules overlap.
  
  Facts: the company is [Company Name] and the planned service is [Service Description]; the model is [Model Origin] trained on [Training Data Sources]; the users are [User Groups]; the operating entity in China is [Operating Entity]; the data processed includes [Data Categories]; the hosting and any cross-border flows are [Hosting and Data Flows]; the content the service can generate is [Content Types]; and the planned launch date is [Launch Date].
  
  Deliverables:
  1. A threshold section determining which regimes apply to this service and why.
  2. A checklist of licences, registrations and filings needed before launch, with the authority and the sequence.
  3. A checklist of model and training data requirements, including lawful source, intellectual property and quality controls.
  4. A checklist of content security duties, including review mechanisms, complaint handling and generated content labelling.
  5. A checklist of personal information duties, including legal basis, notice, consent, minors and user rights.
  6. A checklist of data security and cross-border transfer duties, with the mechanism needed for each flow.
  7. A checklist of corporate, foreign investment and telecom licensing points for the operating entity.
  8. A table of open items rated blocking, high or manageable, with an owner and a suggested date.
  9. A list of facts that must be confirmed because the answer changes the applicable regime.
  
  Mandatory rules: never invent a statute, regulation, measure name, article number, case name or citation, and cite the named source for every rule, threshold or deadline or mark it unverified and flag it for Chinese counsel. Do not treat a draft or proposed measure as in force; label its status. Do not assume that a foreign model or data centre outside China is lawful for this service without a cited basis. Mark every item as met, not met or unknown, and never as met where facts are missing.

• Saudi AI Services Agreement Negotiation MemoMemoEnterpriseNegotiate / Counter

  You are a lawyer on the internal legal team of a technology provider negotiating an artificial intelligence services agreement with a Saudi customer, and you will write a concise negotiation memo for the deal team. Reflect the Saudi framework in general terms, including the personal data protection regime and the data and AI authority, cloud and data hosting rules, requirements for government and critical sector customers, localisation and national security considerations, and the rules on liability, penalties and governing law in a Saudi contract. Distinguish a private customer from a government-related one.
  
  Facts: our company is [Company Name] and the customer is [Customer Name] in [Customer Sector]; the services are [AI Services]; the data involved is [Data Involved]; the hosting is planned in [Hosting Location]; the deal value is [Contract Value]; the customer's main demands are [Customer Demands]; our non-negotiables are [Our Red Lines]; and the signing target is [Signing Date].
  
  Deliverables:
  1. A top-line view of the deal risk and the three issues most likely to block signature.
  2. A table of the regulatory conditions the agreement must satisfy, with the party responsible for each.
  3. A table for data, hosting and localisation terms with our opening position, fallback and walk-away point.
  4. A table for intellectual property in models, outputs and customer data with the same three positions.
  5. A table for liability, indemnity, service levels and penalty terms, checked against Saudi limits on exclusion and on agreed damages.
  6. A table for AI-specific warranties, explainability, audit rights and human oversight.
  7. A note on governing law, forum and arbitration and the enforceability of a foreign choice in the Kingdom.
  8. Sequencing advice for the negotiation, saying which issues to trade and which to settle first.
  9. A list of facts and regulatory points to confirm with Saudi counsel before the next round.
  
  Mandatory rules: never invent a statute, regulation, authority rule, article number, case name or citation, and cite the named source for every rule, rate or deadline or mark it unverified and flag it for Saudi counsel. Do not state that a term is enforceable in the Kingdom unless a source supports it. Do not treat a draft guideline as binding; label its status. Keep the memo short and use tables for positions only.

• AI Service Agreement Clause Review for TunisiaClause / ScheduleExpertReview / Redline

  You are reviewing, for the in-house legal team of a Tunisian company, the key clauses of an agreement under which a foreign vendor supplies an artificial intelligence service, and you must be concise and return replacement wording where a clause is weak.
  
  Facts: the customer is [Customer], the vendor is [Vendor and country], the service does [AI service description], the data it will process includes [Types of data including personal data], the governing law and forum are [Governing law and forum], and the clauses provided are [Clauses to review].
  
  Deliverables:
  1. A review of the clause on use of the customer's data, including whether the vendor may train its models on it, and replacement wording.
  2. A review of the personal data clauses, including roles, security, breach notice and transfer outside Tunisia, and replacement wording.
  3. A review of the clauses on output ownership, intellectual property warranties and indemnity for infringing output, and replacement wording.
  4. A review of the limits on liability, service levels and audit rights, with the figures that look inadequate for the risk and a suggested alternative.
  5. A short list of the points that depend on Tunisian rules and need local confirmation before signature.
  
  Mandatory rules: never invent a statute, article number, case name or citation, and cite the named source for every rule, notice period or deadline or mark it unverified and flag it for local counsel. Quote each clause you comment on instead of paraphrasing it. Do not say that a transfer of personal data abroad is permitted without stating whether authorisation from the national data protection authority is required, or marking that point for confirmation.

• Plain-Language Letter to a Moroccan Regulator About an AI ToolLetter / EmailStandardClient Communications

  You are drafting a letter in plain language from a company to a regulator in Morocco, informing it about an automated decision tool the company plans to use with customers and asking whether any approval or notice is needed. The letter must be easy to read and must not overstate what the tool does.
  
  Facts: the company is [Company Name], active in [Sector]; the regulator is [Regulator Name]; the tool is used to [Tool Purpose] and affects [Affected People]; it uses these data and signals [Data Used]; a person reviews its output in these cases [Human Review]; the planned launch date is [Launch Date].
  
  Deliverables:
  1. A letter of no more than one page that opens with the question being asked.
  2. A short description of the tool in everyday words, saying what it decides, what it only suggests and how people can challenge a result.
  3. A list of the three questions put to the regulator, numbered, including whether any filing or approval is required before launch.
  4. A closing line offering a meeting and naming a contact person.
  
  Mandatory rules: never invent a statute, regulation, article number or citation, and never name a specific legal duty unless it comes from a source the user supplies; cite the named source for every rule mentioned, or mark it unverified and flag it for local counsel in Morocco. Do not claim the tool is lawful, accurate or free of bias. Do not admit any breach. Write in the language the regulator uses and keep sentences short. This letter is a draft for review by counsel and is not legal advice on the company's specific situation.

• Board Playbook for Launching an AI Product in IndiaPlaybookExpertStrategy / Scenario

  Prepare an analytical strategy playbook for the board of [Company Name] on launching an AI product in India, setting out the regulatory exposure, the choices open to the company and a recommended path. The board decides on risk and sequencing, so frame every point as a choice with a consequence.
  
  Facts: the product is [Product Description]; it processes [Data Used]; its users and markets are [Users and Markets]; it is deployed as [Deployment Model] and relies on [Third-Party Models]; the board's risk appetite is [Risk Appetite]; the launch decision is due by [Decision Date].
  
  Deliverables:
  1. A map of the regimes likely to apply to the product: data protection, sector regulators, consumer protection, intellectual property, and rules on content or intermediaries, each with a one-line reason.
  2. An assessment of the data protection position, covering the lawful basis for processing, notice and consent, children's data, cross-border transfer and the duties of a data fiduciary, with the points that need confirmation.
  3. A risk table for model behaviour, covering accuracy, bias, harmful output, intellectual property in training data and outputs, and liability when a third-party model is used.
  4. The contract positions to secure from vendors and customers: warranties, indemnities, audit rights, data use limits and incident notification.
  5. Three launch options, such as full launch, limited pilot and delayed launch, with the regulatory exposure, cost and time of each.
  6. A recommended path with decision gates, the owner of each and the evidence the board should see before moving on.
  7. A one-page board summary with the recommendation, the main risk and what would change it.
  
  Mandatory rules: never invent a statute, article number, case name or citation; cite the named source for every rule, rate or deadline or mark it unverified and flag it for local counsel. Where you refer to the Digital Personal Data Protection Act, 2023, say which of its provisions depend on rules or notifications not yet confirmed, and do not describe their content from memory. Treat draft rules, guidelines and advisories as unverified until a named source is given. Do not state penalties as figures unless you cite them. This playbook is not advice on the company's specific product.

• Board FAQ on Employee Use of Generative AI ToolsQ&A / ExplainerExpertSummarize / Extract

  Summarise for the board of [Company Name] the legal exposure created when employees use generative AI tools at work, as a concise FAQ under the law of [Jurisdiction]. Each answer should be two or three sentences and end with what the board should ask management or approve.
  
  Facts: the tools in use are [AI Tools in Use]; they are used in [Business Functions]; the data fed into them includes [Data Categories]; the existing policies are [Existing Policies].
  
  Deliverables:
  1. Whether the company is responsible for what employees do with these tools, and for what they publish or send to clients as a result.
  2. What the law of [Jurisdiction] says in general about personal data, confidential information and trade secrets going into an external tool.
  3. Who owns the output, whether it can be protected, and the risk that it reproduces someone else's protected material.
  4. What the company owes clients and regulators when AI is used in work for them, including disclosure, supervision and professional duties.
  5. The practical controls a board should expect: policy, approved tools, training, logging and vendor terms.
  6. The five questions the board should put to management at the next meeting.
  
  Mandatory rules: never invent a statute, article number, case name or citation; cite the named source for every rule, rate or deadline or mark it unverified and flag it for local counsel. If [Jurisdiction] has no specific rule on this subject, say so and describe the general law that applies instead, without implying a dedicated regime. Treat any draft law or guidance as unverified until a named source is given. Keep to what the board must decide; do not recommend a particular vendor. This FAQ is general information for the board and not advice on the company's specific situation.

Cerca tra tutti i 631 prompt