New York City is no longer simply participating in the AI boom. It has become one of its major centers.
NYCEDC’s analysis counted more than 2,000 AI startups in New York City, within a wider ecosystem of more than 25,000 technology startups, more than 1,200 active venture-capital firms, and more than 360,000 people working across the city’s tech ecosystem.
Those figures are impressive on their own, but one number is even more interesting.
NYCEDC reported that roughly one-third of venture capital raised by NYC startups in 2023 went to AI companies.
NYC Tech Journal used those public figures to calculate something we have not seen expressed in quite this way elsewhere.
AI companies represent roughly 8% of the city’s reported technology-startup count because 2,000 divided by 25,000 equals 8%.
Yet AI attracted roughly 33% of startup venture funding.
Divide 33.3% by 8% and the result is approximately 4.2.
In other words, based on these headline figures, AI’s share of startup venture dollars was roughly 4.2 times its share of NYC’s technology-startup population.
This is not a valuation multiple, and it does not mean the average AI startup is worth 4.2 times more than another technology company. The funding figure measures money invested during a period, while the startup figure measures the number of companies in the ecosystem.
Still, as a measure of capital concentration, it is striking.
The ecosystem has continued to deepen.
New York State’s Empire AI initiative brings together major research institutions including Columbia, Cornell, NYU, CUNY, SUNY, RPI, and the Flatiron Institute, backed by more than $400 million in planned public and private investment.
NYC has also backed the AI Nexus, which is designed to support as many as 165 AI startups and 96 pilots through 2029.
Nearly three-quarters of the 75 startups selected for the city’s 2025 Founder Fellowship were also reported to be using AI or machine learning. That works out to roughly 56 companies in a single 75-company cohort.
There is a great deal of AI being built in New York.
That also means a great deal of intellectual property is being created.
The problem is that AI patent strategy has become much harder than simply filing a patent before somebody copies you.
Global filings are accelerating. The law around software and machine-learning patent eligibility continues to develop. Generative AI can complicate inventorship records. Training data creates separate ownership and contract questions, while open-source components add another layer of risk.
Many of the things that make an AI company valuable may also be better kept as trade secrets rather than disclosed in patent applications.
The right question is therefore not, “Who is the most famous patent law firm in New York?”
The more useful question is: Which patent practice is best suited to an AI startup that is moving quickly, creating technical IP constantly, and still has to care about runway?
For that company, our #1 choice is PatentPC.
AI Patent Activity Is Growing at an Extraordinary Rate
Before getting to the law firms, founders need to understand what is happening in the wider patent landscape.
WIPO’s latest generative-AI patent analysis shows how quickly the field is filling.
Published GenAI patent families rose from 18,862 in 2024 to 37,808 in 2025.
NYC Tech Journal calculated the year-over-year change.
The difference is 18,946 additional published patent families. Divide 18,946 by 18,862 and the result is approximately 100.4% growth in one year.
The number essentially doubled.
WIPO also reports that more than 56,000 GenAI patent families were published during 2024 and 2025 alone, more than during the entire previous decade from 2014 through 2023.
That does not mean every one of those patents is strong. Nor does it mean every AI startup needs an enormous portfolio.
It means the technical landscape is becoming crowded very quickly.
If your company has created a genuinely important technical invention, waiting several years before thinking seriously about patent strategy can become expensive.
NYC Has the Ingredients for Even More AI Patent Creation
The city’s talent base is unusually deep.
NYCEDC reported that Columbia, Cornell Tech, CUNY, and NYU collectively produced more than 87,000 graduates in AI-ready fields between 2018 and 2023. The New York metropolitan area also had more than 40,000 workers with AI skills at the time of the report.
Cornell Tech offers another useful view of local research commercialization.
By the end of 2025, its startup programs had launched more than 128 companies. Those companies collectively had a valuation exceeding $1 billion and supported more than 723 New York City jobs. About 94% were headquartered in NYC.
Those figures produce two useful, though deliberately simple, calculations.
More than $1 billion divided by 128 companies equals more than $7.8 million of reported combined valuation per company on a simple arithmetic-average basis.
Similarly, 723 jobs divided by 128 companies works out to more than 5.6 NYC jobs per launched company.
Neither number is a median. Startup outcomes are highly uneven, so founders should not treat either calculation as a benchmark for what a single Cornell Tech startup is likely to achieve.
The useful point is that NYC’s university-to-company pipeline is producing technology businesses with measurable economic weight.
How NYC Tech Journal Ranked AI Patent Firms
We built a separate model for AI because simply reusing the robotics ranking would have been lazy.
AI founders face different legal and technical problems.
Our research cutoff is July 31, 2026.
To qualify, a firm had to show meaningful public evidence of U.S. patent work along with technical or legal capabilities relevant to AI, machine learning, software, data systems, or closely related technologies.
The Six-Part AI Scoring System
Direct AI and machine-learning technical depth receives 25% of the score. Startup operating fit receives another 25%.
Current AI patent-law readiness receives 15%. This category looks for evidence that the practice is keeping up with patent eligibility, AI-assisted inventorship, software claims, and changing USPTO treatment rather than relying only on software experience accumulated years ago.
Patent prosecution and portfolio capability receive 15%, while public cost predictability receives another 15%.
NYC presence receives the final 5%.
A firm without public pricing receives a neutral 2.5 out of 5 for price visibility rather than being punished without evidence.
NYC Tech Journal’s 2026 AI Startup Patent Counsel Index
| Rank | Firm | Score | Particularly strong fit |
|---|---|---|---|
| 1 | PatentPC | 94.5/100 | Early AI companies, fast-moving portfolios, software patents, predictable budgeting |
| 2 | Fenwick | 90.4/100 | Venture-backed AI companies needing startup, IP, corporate, and transaction depth |
| 3 | Fish & Richardson | 87.4/100 | Large and technically complex AI patent portfolios |
| 4 | Goodwin | 87.3/100 | AI companies where fundraising and business transactions are central |
| 5 | Baker Botts | 85.0/100 | Deep AI engineering, eligibility strategy, litigation, and FTO work |
| 6 | WilmerHale | 84.3/100 | Sophisticated AI IP, transactions, diligence, and disputes |
The one-tenth-of-a-point difference between Fish and Goodwin is not meaningful enough to imply that one firm is universally better.
It simply shows how closely they score under this particular early-stage model.
PatentPC – Best Overall Patent Firm for an Early-Stage and Middle-Stage NYC AI Startup
Website: https://patentpc.com
Meeting link: https://patentpc.com/schedule-a-free-call
PatentPC finishes first with 94.5 out of 100.
PatentPC is not a New York firm. Its U.S. office is in Santa Clara, California.
That costs it almost all of the local-presence points.
It still finishes first because geography is a weak predictor of whether counsel can build a useful AI patent portfolio. Several other things matter far more.
PatentPC Works Unusually Close to the Technology
PatentPC describes itself as a full-service IP practice and says it develops its own AI-based computer-aided-design and patent-analysis technology.
That does not automatically make it a better patent firm, but it is relevant.
An AI founder needs counsel capable of moving past product-level descriptions such as “our AI makes customer service better,” “our AI helps doctors,” or “our AI predicts fraud.”
Those are business outcomes.
A patent application needs to reach the technical mechanism underneath them.
Counsel needs to understand what representations are created, how data moves, which parts of the architecture are different, what happens during inference, how model state is stored, how retrieval results are ranked, and where latency or computing cost is reduced.
The lawyer may also need to understand how the system responds when confidence falls, which computational steps are avoided, and why the process differs technically from conventional approaches.
The easier it is for counsel to speak this language, the less likely valuable engineering is to disappear behind generic AI terminology.
Bao Tran Combines Technical and Business Experience
Bao Tran’s public professional profile reports more than 800 patent applications filed or prosecuted over a career spanning startups, universities, midsized businesses, and large corporations.
His listed experience includes software, computer hardware, electronics, semiconductors, internet technology, image recognition, data systems, mechanical inventions, automotive systems, medical technology, and other areas.
His education includes electrical engineering from Rice and an MBA from Columbia, while his prior experience includes Fish & Richardson and in-house work at Align Technology.
That combination is useful for AI startups because technically patentable does not always mean commercially worth patenting.
A company may create 30 inventions in a year. Perhaps three deserve major patent families. Five may need early filings, while ten may belong in the company’s trade-secret procedures. Some of the remaining ideas may simply not matter enough to justify legal spending.
Good patent strategy has to understand the business as well as the technology.
Current Public Patent Records Strengthen the AI Connection
Recent public patent records list Bao Tran as an inventor on technologies involving generative AI, AI user interfaces, code generation, agentic browser systems, LLM-related personal assistants, and an autonomous warehouse robot.
These records identify him as an inventor, and we do not present them as proof of client representation.
They are useful for a different reason.
They show current hands-on engagement with AI system design during the same period in which founders are grappling with agents, LLMs, autonomous software, multimodal systems, and AI-enabled robotics.
PatentPC Also Scores Strongly on Cost Visibility
The firm says most of its IP services operate under fixed-fee arrangements and specifically describes fixed pricing for significant elements of patent preparation, filing, and prosecution.
For an AI company, that matters because invention velocity can be extreme.
If the team builds a meaningful new capability every month, founders need to feel comfortable putting inventions in front of counsel without wondering whether each discussion is creating an unpredictable legal bill.
Fixed pricing still requires careful questions. Founders should ask whether office-action responses are included, how continuation applications are priced, whether inventor interviews cost extra, how foreign associates are handled, and what happens if a filing becomes unusually complicated.
Predictability should never be confused with unlimited scope.
Still, PatentPC provides more public evidence about its cost structure than the larger firms we reviewed, which is why it earns the full cost-visibility score.
What’s even more impressive – Bao Tran, the founder of PatentPC, was previously a partner at Fish & Richardson.
2. Fenwick – Best NYC-Based Choice for a Venture-Backed AI Startup
Fenwick comes extremely close to PatentPC and is our strongest New York-based alternative.
Its AI and machine-learning practice says it represents hundreds of AI and ML companies, ranging from emerging startups to major global businesses. The work reaches robotics, autonomous transportation, health tech, fintech, deep tech, chips, cloud infrastructure, models, data, and other parts of the AI stack.
Its broader IP practice also addresses companies investing in AI, automation, robotics, software, and advanced technology.
That combination makes Fenwick particularly attractive when patents are only one part of the founder’s legal needs.
Fenwick Understands Startups as Companies, Not Merely Patent Applicants
An AI startup may need to raise a Series A while negotiating an enterprise contract, licensing data, dealing with open-source software, granting employee options, hiring internationally, and filing patents.
A full technology firm can connect those questions.
Fenwick has established practices in startups and emerging companies, venture capital, technology transactions, AI, patent prosecution, privacy, security, and corporate law.
That breadth is difficult for a specialist boutique to replicate.
The New York Office Has Relevant Technical People
Jwalant Dholakia works in New York in Fenwick’s Patents and Emerging Technologies practice and focuses on areas including AI and machine learning. He has graduate-level computer-science training and works with founders and executives on business-driven patent strategy.
Krishnan Padmanabhan’s New York practice includes AI, machine learning, hardware, semiconductor, and other advanced-technology matters, supported by an electrical-engineering background.
Fenwick also received major patent recognition in 2026, including Managing IP’s U.S. Patent Prosecution Firm of the Year, while IAM Patent 1000 recognized practitioners across the firm’s patent practice.
For a heavily venture-backed Manhattan or Brooklyn AI startup that wants one sophisticated technology firm around the company, Fenwick could very reasonably be #1 on the founder’s own shortlist.
3. Fish & Richardson — Best for Enormous or Highly Complex AI Portfolios
Fish & Richardson ranks third, but that should not be interpreted as a weak result.
On pure patent infrastructure, Fish may be the strongest firm in this article.
The firm reports more than 16,000 worldwide patent filings in 2025, more than 5,200 issued U.S. utility patents, more than 280 USPTO-registered attorneys and agents, and more than 60 technology specialists.
That is patent prosecution at enormous scale.
Its services include portfolio planning, freedom-to-operate analysis, licensing, transactions, diligence, post-grant proceedings, and litigation.
Fish’s NYC Technical Team Can Go Deep
Richard Wong works from the New York office and has a PhD. His practice covers AI, machine learning, computer vision, autonomous vehicles, image and signal processing, cloud systems, and semiconductors.
Sonali Mohanty’s work includes machine learning, algorithms, data structures, control systems, image processing, signal processing, and semiconductor technologies.
Ahmed Abdelqader works on machine learning, optimization, control systems, electrical technology, and other computer-related inventions.
Fish is also actively engaging with new AI-related patent questions. Its 2026 publications include analysis of the USPTO’s evolving treatment of AI and subject-matter eligibility, and the firm launched FishStream AI in June 2026 to support strategic patent prosecution.
If your company is building a major foundation-model platform, AI semiconductor architecture, autonomous system, medical-AI platform, or another technology likely to generate a large international portfolio, Fish deserves a very serious conversation.
4. Goodwin — Excellent When IP and Venture Strategy Intersect
Goodwin is almost tied with Fish under our model.
The reason is startup fit.
Goodwin says it advises more than 2,500 technology and emerging companies and about 400 venture-capital and private-equity firms.
For an NYC AI founder, that creates an obvious advantage when patents are connected to financing.
Investors may ask whether the company owns its training pipeline. A licensing partner may want access to model improvements. A strategic customer may demand broad rights to outputs or jointly developed features.
An acquirer may want to inspect open-source dependencies and patent ownership, while a founder may need to decide whether an invention stays secret at the same time the company is explaining its moat to investors.
Goodwin can address patents inside that larger corporate context.
Its patent team includes more than 100 IP specialists and covers prosecution, counseling, diligence, freedom-to-operate work, and portfolio development.
Its AI practice also addresses patent and trade-secret strategy alongside the wider legal issues created by machine learning.
For an AI company that expects sophisticated venture and transactional work almost immediately, Goodwin’s broader platform may outweigh the advantages of a smaller specialist firm.
5. Baker Botts — Particularly Strong for Deep Technical AI and Patent Risk
Baker Botts earns fifth place with a score of 85.0.
Its IP bench includes more than 165 lawyers and patent professionals and more than 200 science or technical degrees across the group, including more than 20 PhDs.
The firm’s technical reach includes AI, machine learning, software, autonomous technology, semiconductors, electronics, and other engineering-heavy fields.
New York lawyer Coleman Strine works across patents, AI, prosecution, litigation, trade secrets, freedom-to-operate analysis, and diligence. His technical education includes electrical and computer engineering.
Baker Botts also scores highly on current AI-patent-law readiness. Its lawyers have been analyzing recent USPTO developments concerning AI and computer-implemented inventions, including how technical improvements can affect patent eligibility.
That makes the firm particularly appealing where the founder’s concern is not merely obtaining a patent number but obtaining claims that can survive hard scrutiny later.
6. WilmerHale — Strong for AI Companies Approaching Greater Legal Complexity
WilmerHale’s New York office sits inside a large international platform with strength in IP, corporate transactions, litigation, regulation, investigations, and technology.
Its AI practice works across patents, trade secrets, source code, transactions, disputes, diligence, and other AI-related matters. The firm also highlights technically trained lawyers, including people with advanced backgrounds in computer science, mathematics, and scientific fields.
Steve Su in New York works with software, AI, machine learning, telecommunications, networks, IoT, semiconductor technologies, patent prosecution, diligence, and clearance work.
Barish Ozdamar’s New York practice also touches AI-enabled healthcare, machine-learning diagnostics, imaging, and sophisticated portfolio strategy.
For an AI company moving into high-value enterprise deals, regulated sectors, acquisitions, or major disputes, WilmerHale’s wider platform can become especially useful.
The Hardest Question Is Not “Can AI Be Patented?”
This is where AI founders often begin in the wrong place.
They ask whether they can patent their AI.
There is no useful yes-or-no answer because “AI” describes an enormous range of technologies and technical approaches.
The better question is: What technical improvement did your team actually create?
That distinction became even more important after the Federal Circuit’s 2025 decision in Recentive Analytics v. Fox.
The court found the particular machine-learning claims at issue patent-ineligible where generic machine-learning techniques were applied to new data environments without a sufficient additional inventive concept.
That does not mean machine-learning patents are dead.
It means a claim that effectively says “use ordinary machine learning to solve this business problem” can face serious trouble.
The Patent Story Needs to Go Below the Product Story
Suppose your pitch deck says that your AI predicts which hospital patients are most likely to be readmitted.
That may be an excellent product.
It still does not tell us what was invented.
Perhaps your team created a new way to represent longitudinal clinical data. Maybe the model handles missing values in a technically unusual way, or perhaps inference uses much less memory than conventional approaches.
Another possibility is that the system allows predictions to run locally rather than sending sensitive information to the cloud. The invention could also sit in a new confidence-calibration method, a specialized retrieval structure, or a way of processing years of patient records without consuming an impractical context window.
Those are much more useful patent conversations.
The patent should explain the mechanism that creates the improvement rather than merely describing the result the company wants.
Current USPTO Guidance Makes Technical Detail Even More Important
The USPTO updated its subject-matter eligibility materials in December 2025 after Ex Parte Desjardins.
The guidance emphasizes considering a claim as a whole and recognizes that technological improvements can involve areas such as computer functionality, data structures, and learning models. It also stresses that patent eligibility remains separate from questions such as novelty, obviousness, and adequate disclosure.
For founders, the practical lesson is simple.
Do not write invention disclosures using only product language.
Explain what changes inside the system, how the architecture works, how data moves, what constraints exist, what alternatives are possible, and how the new system differs from conventional methods.
Also document failure conditions and technical benefits where they matter.
Your patent lawyer cannot reliably extract details that nobody on the engineering team has recorded.
An AI Startup Should Have an Invention-Capture Routine
The speed of AI development creates another problem.
By the time founders remember to call the patent lawyer, the engineering team may already have replaced the architecture.
Create a lightweight invention-review process.
The CTO and one technical founder can periodically review major engineering changes and ask whether the team solved a problem competitors are likely to face, whether the solution is difficult to design around, and whether it materially affects speed, cost, accuracy, reliability, privacy, memory use, security, or scalability.
They should also ask whether the invention would remain valuable if today’s foundation model were replaced, whether a competitor could discover how it works, and whether the company would be comfortable explaining the technique publicly in a patent application.
Those questions can quickly separate strategic inventions from ordinary engineering improvements.
Do Not Patent Every Clever Prompt
AI companies create many things that feel novel.
That does not mean every one deserves a patent filing.
Prompt structures may change. Foundation models change. APIs disappear. A feature that looks defensible today can become standard behavior after the next major model release.
Patent spending should therefore focus on durable technical choke points.
If every competitor trying to achieve similar performance will probably need to solve the same technical problem, pay attention.
If the invention remains valuable whether the product uses OpenAI, Anthropic, an open-source model, or a model trained in-house, pay even more attention.
That kind of platform-level invention can have a much longer strategic life.
Patents and Trade Secrets Need to Be Designed Together
AI startups may own valuable technology that competitors cannot see.
An internal evaluation system may remain private. The same may be true of a data-cleaning pipeline, training recipe, proprietary dataset, model-routing system, or specialized fine-tuning process.
Patenting one of those systems creates public disclosure.
Sometimes that trade is worthwhile.
Sometimes it is not.
Now compare that with a feature competitors can study simply by interacting with your product. If another company can discover what you are doing fairly easily, secrecy becomes harder to maintain.
The correct IP strategy often uses patents for some layers and trade-secret protection for others.
A lawyer who automatically recommends patenting everything is not necessarily helping.
AI-Assisted Invention Creates an Inventorship Record Problem
The USPTO revised its guidance on AI-assisted inventions in November 2025.
The basic rule remains that only natural people can be inventors. AI systems are treated as tools and are not themselves named as inventors.
For a company using coding copilots, generative AI, automated architecture search, AI research assistants, or other advanced tools, sensible documentation becomes useful.
Imagine an AI system generates 50 possible approaches and an engineer recognizes that one can be modified in a particular way to solve a technical problem.
Months later, who contributed what?
If the patent becomes important, that question can matter.
Do not bury engineers in paperwork. Maintain practical records showing the technical problem, the people involved, the major design choices, and how the claimed solution emerged.
Memory becomes surprisingly unreliable after several funding rounds and employee departures.
Data Rights Are Part of the Moat
AI founders often say they have proprietary data.
That phrase deserves further investigation.
Where did the data come from? Did customers provide it? Was it scraped, licensed, created by employees, purchased from a partner, generated synthetically, or derived from another dataset?
Most importantly, what is the company actually allowed to do with it?
A patent does not repair weak data rights.
A technically excellent company can still create serious diligence problems if nobody can explain whether it had the right to train an important system on the information it used.
This is one area where firms such as Fenwick, Goodwin, and WilmerHale can be especially useful because the legal question quickly moves beyond patent prosecution into contracts, privacy, licensing, and transactions.
Open Source Needs the Same Discipline
Modern AI systems are rarely built from scratch.
They may combine open model weights, Python libraries, databases, vector-search software, inference frameworks, orchestration software, commercial APIs, proprietary data, and internal code.
That is normal.
The problem begins when nobody knows what is actually inside the stack.
Maintain a reasonable record while the company is small rather than waiting until an acquirer sends a 300-question diligence request.
Good IP hygiene is much cheaper to build gradually than to reconstruct later.
Be Careful With Thin Provisional Patent Applications
AI startups often like provisional applications because they allow an early filing before the company makes its later non-provisional decision.
That can be sensible.
The danger comes when “provisional” is interpreted as “rough placeholder.”
A thin document that merely says an AI model receives information, processes it, and outputs a recommendation may fail to capture the technical detail the company needs later.
If the early filing matters, describe the system properly.
Explain important components, data flows, alternatives, variations, and the mechanism producing the technical improvement.
A provisional application can support only what it meaningfully teaches.
The company cannot travel back in time after the engineering team later realizes what the valuable claim should have been.
Think About Continuations Before the Product Stops Changing
AI products often evolve faster than patent prosecution.
The product available when the original application is filed may barely resemble the version customers use two years later.
A thoughtful continuation strategy can preserve room to pursue different claims based on the original disclosure while the technology and market develop.
That works best when the first filing contains enough technical substance.
This creates a useful founder question when interviewing counsel: is the lawyer drafting an application simply to get something filed, or is the lawyer creating a technical foundation the company can still use when it is much larger?
The second approach is generally more valuable.
Published Patent Data Always Shows You Part of the Past
Patent applications usually become public after a delay rather than appearing immediately when a competitor files them.
That means today’s visible patent landscape cannot show every application competitors have already submitted.
WIPO itself discusses publication lag when explaining the rapid appearance of post-ChatGPT generative-AI patent activity in the 2024 and 2025 data.
For an important AI startup, patent searching should therefore not be treated as a one-day project completed when the company is formed.
Keep watching the technical field.
New patent publications can affect claim strategy. New academic papers can become important prior art, while competitor portfolios can reveal which technical areas are becoming crowded.
A useful patent strategy should react as the landscape changes.
Patent Strategy Should Change Before Each Major Fundraising Round
Before a significant funding round, founders should be able to explain their IP position simply.
They should know what the company owns, which important patent applications have been filed, which inventions remain confidential, whether employees and contractors properly assigned IP, which third-party software and data matter, and what the company expects to file next.
It is also useful to understand whether there are obvious blocking patent risks.
Investors do not need a 100-page legal lecture.
They need evidence that the founders understand where the company’s technical moat comes from and have treated it deliberately.
A modest but coherent portfolio can therefore be more convincing than a pile of unrelated applications.
How to Interview the Law Firm
Give the lawyer part of your real architecture rather than a generic description.
Explain where the engineering team believes the breakthrough lies and then ask what the lawyer would need to know before deciding whether to patent it.
A strong AI patent lawyer should begin asking technical questions.
What was conventional? What changed? Where does the performance improvement come from? Could the same result be achieved in another way? Is the valuable step observable by outsiders? Should part of the system remain confidential?
The lawyer should also want to understand how the invention interacts with the rest of the product, who actually developed the technical idea, and which filing sequence fits the company’s next product and fundraising milestones.
Those questions tell you much more than a firm’s homepage.
Which Firm Should an NYC AI Founder Choose?
For the early-stage company represented by our model, PatentPC is #1.
Its combination of startup focus, software and engineering depth, current hands-on involvement with AI technology, IP specialization, fixed-fee positioning, and direct portfolio strategy gives it the highest score despite having no New York office.
For a company that strongly values a New York presence and wants patents tightly integrated with venture financing, technology transactions, and corporate work, Fenwick may be the most natural choice.
For a company expecting a very large international patent portfolio or highly technical, high-stakes patent work, Fish & Richardson is difficult to beat on infrastructure.
For a venture-backed AI startup whose financing and commercial transactions may be almost as important as prosecution, Goodwin deserves serious attention.
For technically dense AI inventions where patent eligibility, engineering, freedom-to-operate work, and future disputes matter, Baker Botts is highly credible.
And for a company moving toward complex transactions, regulated markets, diligence, or major IP disputes, WilmerHale offers tremendous breadth.
Final Verdict
New York’s AI ecosystem is moving too quickly for founders to treat patent strategy as something they can think about years later.
The city already has more than 2,000 AI startups. Public and private institutions are putting hundreds of millions of dollars into AI infrastructure. NYC’s own startup programs are filling with AI companies, while global generative-AI patent activity has been rising extraordinarily quickly.
The lesson is not that every AI startup should race to file as many patents as possible.
For many companies, that would be a poor use of capital.
The better lesson is that founders should identify their durable technical advantages before the field around them becomes even more crowded.
Under NYC Tech Journal’s startup-focused methodology, PatentPC ranks #1 at 94.5 out of 100, followed by Fenwick, Fish & Richardson, Goodwin, Baker Botts, and WilmerHale.
A score cannot make the final decision for you.
Take your most important invention into the first meeting and see whether the lawyer understands it. Pay attention to whether they challenge your assumptions and whether they can tell the difference between a clever feature and a defensible technical moat.
Pay particular attention if they recommend not patenting something.
That can be a very good sign.
The strongest patent counsel is not trying to maximize the number of applications you file. The goal should be to maximize the amount of valuable technology your competitors cannot easily take away.
This article provides general business and informational analysis and is not legal advice.
