Paul Starrett
Tax and legal questions about technology turn on what the technology really does. That’s where I work.
Advisory by default, with legal representation where a matter calls for it.
Attorney and technologist. Twenty-five years at the intersection of law and technology, specializing in AI and related technologies since 2013.
Taxation of technology
How technology is taxed depends on what was built, how, and where. Across the AI stack, I establish the technical facts each question turns on, alongside your tax advisors.
Not every question needs me. For routine R&D credit studies, automated tools and credit firms are often the most cost-effective choice, and I can help you choose among them. My work earns its cost when the stakes or the technical complexity are high: AI-heavy claims, examinations, large or novel positions, transactions, and technology moving between entities.
Cloud infrastructure and compute
- R&D credit treatment of cloud and compute costs, and the split between development and production
- Section 174 and 174A treatment of compute spent on development
- How cloud and digital-content transactions are classified, and where their income is sourced
- Sales and use tax on cloud services; data-center incentives
Data
- Whether data acquisition and preparation costs are deducted or capitalized
- Whether data licensing is a royalty, a sale, or a service
- Who develops and controls data assets across entities (DEMPE)
Model development (training, fine-tuning, evaluation)
- Which experiments qualify for federal and California R&D credits
- Section 174A expensing for domestic research; Section 174 amortization for foreign research
- Contract and funded research when development is outsourced
- Who develops, enhances, and controls the model, and where (DEMPE; cost sharing)
Software and applications
- Internal-use software rules for the R&D credit
- Whether software and digital-content income is a license, a sale, or a service
- Sales tax on software and SaaS; digital services taxes
Deals and IP moves
- What the AI assets actually are in an acquisition, and how they’re allocated
- Diligence on past credits and Section 174 positions
- Technology and IP moving between affiliates
- How R&D credits are treated under the global minimum tax (Pillar Two)
When it matters most: while the work is happening, when the record is built; before you file or set intercompany terms; under examination; and in a transaction.
The same skill, other questions
A record built while the work happens serves more than one purpose.
- AI governance: whether a system does what its policy says it does, documented before a regulator or plaintiff asks
- Disputes and investigations: reconstructing what a system actually did when its outputs are challenged, as counsel, consulting expert, or testifying expert witness
Proving technology work takes three kinds of expertise. They’re usually split across three people; here they’re one.
When those roles are split, the gaps get filled with assumptions, and assumptions are exactly what an examiner, a regulator, or opposing counsel will find.
The technology
Understand how the system is built, trained, and run, well enough to answer the second and third questions, not just the first.
- Former security software engineer, RSA Security
- Former General Counsel and Chief Risk Officer, publicly held AI and data-management company
- Master of Science, Data Science (Predictive Analytics), Northwestern University
- Lead author, Digital Signatures (RSA Press / McGraw-Hill, 2002)
The law
Advise on how the work and its record are set up, so the legal and tax positions hold up when someone tests them.
- Attorney at Law, California
- Master of Laws in Taxation
The people
Documents never hold every fact. Capturing what engineers knew and tried means asking them well, while the work is fresh.
- Certified Fraud Examiner
- Licensed Private Investigator, California
- Author, The Investigative Interviewer's Guidebook, for risk, compliance, and legal professionals (first published 1998)
Alongside the team you already have, in the role the matter needs, at a price agreed up front.
- Alongside your advisors. I supply the technical facts and what they mean for the question at hand; your CPA, transfer pricing advisors, outside counsel, and auditors keep their roles. On an R&D credit, for example, I establish which work qualifies and why, and your CPA computes the credit and prepares the return.
- In the role the matter needs. Advisor, counsel to the company, consulting expert retained by your counsel, or testifying expert witness, and, where facts must be gathered from people, as a licensed private investigator.
- Privilege where it matters. As counsel, or as an expert retained by counsel, the work is protected where the law allows.
- Fixed fees, agreed up front. Scope and price are set before work begins, never as a percentage of a credit.
In practice since 2001: founding chair of the ABA Big Data Committee from 2013 to 2020, more than 45 multimillion-dollar investigation and e-discovery matters for AmLaw 50 firms and Fortune 100 clients, and a contributor to the IAPP's AI Governance certification.
Adjunct faculty at Santa Clara University (School of Law and Graduate School of Engineering) and in the M.S. Data Science program at the University of the Pacific.
Let's talk about what you're building.
Advisory by default, with legal representation where a matter calls for it. The first conversation is direct and confidential.
paul@starrettlaw.com