Services
Full lifecycle legal services for AI industry-chain companies — from setup, compliance and going global to funding and exit, Geason goes deep in one industry.
The foundation from incorporation to daily operations: entity, equity and founder arrangements first, incentives to lock in the team, and contracts, employment and tax to keep operations sound.
Incorporation and Entity Structure
Entity form, holding vehicle, business scope and governance terms decided at incorporation; get them right once and later financing stops being a correction.
Founder Agreements and Equity Structure
How founder equity is split, how vesting and repurchase work, and who owns the IP decide who holds leverage in a round. Fix the structure before the term sheet.
Equity Incentives and Option Pool
An unreserved pool dilutes every founder at the next round, and a casually priced strike makes options worthless. Design pool, vehicle and tax together.
Commercial Contract System
AI contracts turn on three clauses, not templates: liability for model output, whether user input may train the model, and each side's processing role.
Employment Compliance
Staff leaving with model weights and data, and options that never connect to employment, are the two employment failures that recur most in AI companies.
Tax Compliance
R&D super-deduction and high-tech certification are the two tax tools that matter most to AI companies, provided the books and evidence survive audit.
An AI company's assets sit in code, models, data and brand. This block is asset structuring and ownership, not compliance review: algorithm patents, software copyright and trademarks, open-source clearance, trade secrets, training data and personal information.
IP Protection and Layout
Core assets sit in algorithms, code and brand, while patent eligibility and open-source licences are often missed. Clear title during R&D, not before a round.
Trade Secrets and Non-Compete
Model weights, data mixes, prompts and tuning know-how rarely qualify for patents and should not be public. Trade secrecy is the route if grading and logs hold.
Data Compliance and Privacy
Where training data came from, how far de-identification goes, and whether calling an offshore API is a transfer: these three set the compliance baseline.
Cybersecurity Multi-Level Protection
MLPS is not a one-off assessment but a chain of grading, filing, remediation and retesting that restarts when systems change. Most teams meet it at a tender.
Cross-Border Data Transfer
PIPL and GDPR are not a clause-by-clause translation; the work is finding points that look alike but apply differently. Fix the transfer route, then contracts.
Overseas IP Portfolio
Trademarks and patents are territorial: a Chinese registration stops at the border. PCT, Madrid and FTO should follow target markets, not squatting claims.
Domestic market access and regulatory filing: business licensing, algorithm and model filing, AI content labelling, and regulatory investigations — what decides whether a product can launch and charge.
Business Licensing and Market Access
API resale, online data processing and content generation fall outside the licensed scope. Settle ICP/EDI and foreign-investment access before you charge.
Algorithm Registration and LLM Filing
Algorithm registration and LLM filing turn on the security assessment and mechanism documentation; incomplete files come back and changes need a refiling.
AI Content Labelling and Content Safety
Labelling is not a statement that content is AI-made; it keeps content traceable after forwarding and re-editing. Visible and implicit labels cover the chain.
AI Criminal Risk and Investigations
Criminal exposure for AI products usually sits in data acquisition, content output and missing licences. A review before launch costs less than investigating.
Putting the product on the global shelf: cross-border data transfer, the EU AI Act, US and Southeast Asia regulation, export controls, offshore structures, and overseas employment and tax — going global is a chain; one missing link breaks it.
Export Controls and Economic Sanctions
The first gate overseas is not the market but the list: GPU models and weight transfers can be controlled exports. Classify items and screen parties first.
EU AI Act Compliance
The EU AI Act allocates duties by role: providers and deployers carry different lists. Classify risk and settle your role early, before building the system.
US and Southeast Asia AI Regulation
US state laws and Southeast Asian regimes are fragmenting, and copying each wastes effort. Build one baseline, then local adaptation layers in launch order.
Offshore Structure and Red-Chip Setup
Offshore funds want a Cayman structure they can read; domestic regulators want a chain they can look through. Circular 37 and ODI filings must satisfy both.
Overseas Employment and Local Entities
Hiring through an EOR is the usual first step abroad, but the contract is theirs and the management is yours. Joint-employer and PE risk surfaces later.
Overseas Tax and PE Exposure
The underestimated cost of going global is not the tax rate but the years clawed back once a permanent establishment is found. Price repatriation early.
Structuring capital events: equity and debt financing, listing paths and M&A exits — every transaction has its own structure.
Equity Financing and Investor Terms
A term sheet allocates control, not valuation. Liquidation preference, anti-dilution and repurchase matter more than price; model them before you negotiate.
Debt Financing and Public Funding
Financing is not only selling equity. Technology credit, IP pledge, guidance funds and grants raise money without dilution, and are the half often overlooked.
IPO and Capital Markets Path
A-share, Hong Kong and US listings are not three options but three sets of compliance and history demands. The wrong board shows in the first questions.
M&A and Equity Exit
What founders and investors get, in what form and when, depends on cleaning the structure before sale. Drag-along, repurchase and liquidation order come first.
A cross-cutting line, not a stage: disputes can arise at formation, financing, expansion or exit. This group covers handling and loss control once a dispute exists, rather than another lifecycle stage.
Commercial Litigation and Arbitration
AI disputes cluster in four areas: shareholder equity, contract delivery, financing and employment. First ask not litigation or arbitration, but what delivers.
AI Infringement & Unfair Competition
Being sued and suing often happen together: output is called infringing, training data unlawful, and rivals scrape your data. Separate the claim from its cost.