Can AI be Patented?
Why AI and patents matter to innovators
AI is transforming how we invent, optimise and operate. However, innovation often brings questions, especially regarding intellectual property (IP). Can you patent an AI-generated idea? Can AI be an inventor? And what does that mean for start-ups pushing the boundaries of what’s possible?
In this blog, we’ll explain what you need to know about patenting AI-driven inventions, unpack recent legal decisions and explore how to stay protected in today’s AI-powered landscape.
A real-world case study: DABUS and the future of AI and intellectual property
The DABUS case is a landmark in the evolving relationship between artificial intelligence (AI) and intellectual property (IP) rights. It brought into sharp focus the legal boundaries of AI-generated innovation and clarified the role of human inventorship in patent law across the globe.
AI and patent law: The background to the DABUS case
Back in 2018, inventor Dr Stephen Thaler submitted UK patent applications for:
- A food container
- An emergency beacon
He named his Artificial Intelligence system, DABUS, as the sole inventor, arguing that since he owned the AI, he should also own any inventions it generated.
The UK Intellectual Property Office (UKIPO) rejected this view. It maintained that:
- An inventor must be a natural person, and
- Simply owning an AI system didn’t entitle Dr Thaler to own its output
Both applications were refused in 2019.
AI-generated innovation and the legal challenge
Dr Thaler pursued a series of appeals through the UK courts in a bid to overturn the UKIPO’s decision. However, each court upheld the original ruling:
- 2020 – High Court dismissed the appeal ([2020] EWHC 2412 (Pat))
- 2021 – Court of Appeal upheld the decision ([2021] EWCA Civ 1374)
- 2023 – The Supreme Court heard the case and published its final, unanimous decision in December 2023 ([2023] UKSC 49)
Supreme Court ruling: Why AI could not be named as an inventor under UK law
The UK Supreme Court made several key determinations:
- Only a human could be considered an inventor under the Patents Act 1977
- DABUS, being an AI system, was “not a person at all” and therefore ineligible for inventorship
- Owning an AI system did not confer any rights to inventions it produced
- Patent rights could only arise from human inventorship, not tool ownership
- The Court dismissed Dr Thaler’s claim under the Doctrine of Accession, which applies only to tangible property, such as fruit from trees or calves from cows, not to intangible rights like intellectual property.
The Court’s interpretation was rooted in sections 7 and 13 of the Patents Act and reflected the need for consistent application of the law across diverse situations.
AI and intellectual property rights around the world
Thaler’s legal battle wasn’t limited to the UK. He pursued similar applications in multiple jurisdictions, with mixed results:
- UK, EPO, USA, Germany, South Korea – All concluded that AI could not be named as an inventor.
- Australia – Initially accepted DABUS, but that decision was overturned on appeal.
- South Africa became the only country to grant a patent naming DABUS as the inventor.
- European Patent Office – Rejected the DABUS filings but accepted a divisional application naming Dr Thaler himself as the inventor. This divisional application has also been refused based on the inventor claim but this decision is under Appeal and is due to be heard in February 2026.
There appears to be an overall consensus, albeit taking some time to establish in jurisdictions such as Australia, that AI cannot be an inventor.
What the DABUS case means for start-ups using AI in innovation
The implications of the case were significant, especially for businesses and start-ups exploring AI-powered technologies:
- AI couldn’t be credited with inventorship; a human must be named
- IP rights couldn’t be claimed just because someone owned the AI involved
- Early-stage disclosure of AI-generated inventions still posed a risk to novelty
- Filing strategies had to be adapted depending on the jurisdictional law
It underscored the importance of getting inventorship, ownership, and timing right when seeking protection for AI-enabled solutions.
Stratagem’s view on the DABUS ruling and its legacy
The DABUS decision shaped how AI and intellectual property are handled in the UK and beyond. It confirmed a key legal principle: while AI can assist in the innovation process, patent systems still rely on human creativity, authorship, and accountability.
For any business building AI-driven products, this case served as a valuable reminder to:
- Document human contributions carefully
- Secure IP before discussing or demonstrating the invention
- Tailor strategies for each market, particularly when planning international filings
As AI continues to evolve, so too will the legal frameworks that support it. The DABUS case was a milestone, but not the final word on the matter. That’s why staying ahead of changes and working with the right IP partner remains more important than ever.
Practical Aspects of the law today: Where UK and global patent law draws the line
As set out above, at the present time AI cannot be listed as an inventor. Most patent offices, including the UK Intellectual Property Office (UK IPO) and the European Patent Office (EPO), insist on a human inventor.
That means AI can support innovation but can’t own or originate IP rights.
AI and patentability: What can be protected?
While you can’t patent AI, you can patent inventions that use AI if they have a technical effect. These same principles apply to software patents and AI-led solutions.
AI-powered innovations that may be patentable:
- A diagnostic tool that uses AI to interpret complex medical images.
- A control system that boosts energy efficiency through predictive modelling.
Examples unlikely to qualify:
- A basic business method delivered via an AI app.
- A system that changes how information appears without a technical impact.
AI alone doesn’t make an idea patentable. What matters is the result and the technical value it brings.
Why AI can’t be an inventor: from brain to algorithm
Modern AI draws on neural networks, systems modelled on how the human brain works. These concepts have their roots in 19th-century science and gained focus in the mid-20th century. Advances in computing and data science have since enabled these models to be applied in commercial use.
Even the most advanced AI today still follows these foundational principles. Legally, that matters. AI processes data and delivers results. But it doesn’t act with intent, creativity or authorship. That’s why it can’t be named as an inventor.
Global patent law and AI: A mixed landscape
The UK and European offices continue to require human inventors. However, other countries have taken a different view. In 2021, South Africa granted a patent that named an AI system as the inventor. Australian courts briefly supported the idea before reversing their decision.
This legal patchwork adds complexity to global IP strategies. If your business operates internationally, you must consider local laws and remain prepared for future changes.
A timeline of AI and patent law developments
Key milestones in the AI and IP conversation:
- 2019 – DABUS applications were submitted to the UK, the US, and the EPO.
- 2020 – EPO and UK IPO reject AI as an inventor.
- 2021 – South Africa accepts an AI-invented patent. The Australian court briefly does the same.
- 2023 – WIPO and UK IPO launch policy consultations.
- 2024 – UK IPO ends voluntary AI copyright code after failing to reach an agreement.
This evolving picture illustrates why maintaining a flexible and well-informed IP strategy is crucial.
Looking ahead with AI and IP: What might change?
As AI becomes increasingly capable, regulators are under pressure to update intellectual property (IP) frameworks. Ongoing debates include:
- Should AI-assisted inventions still require a human inventor?
- How do we treat copyright in AI-generated content or training data?
- Do current inventorship models work for collaborative human-AI innovation?
These questions could significantly reshape intellectual property (IP) in the next decade. Getting ahead now will help you respond quickly when the landscape shifts.
The role of AI in the patent process
AI isn’t just the subject of patents. It’s also changing how we work with them.
Today, AI-powered tools can help with:
- Speeding up prior art searches.
- Drafting claims using insights from previous filings.
- Spotting patterns in examiner behaviour.
At Stratagem, we use these tools as appropriate for the analysis of published materials. However, we rely on people for claim drafting, strategy, and risk assessment. Human input remains essential to achieving successful outcomes.
Common pitfalls with AI patents: how to avoid them
We often see businesses face similar issues when they seek protection related to AI.
- Saying “It uses AI” without context.
Vague claims won’t convince examiners. Be specific. What kind of AI? What problem does it solve? What’s the measurable impact? - Overstating what AI does.
AI is often just a tool. If your innovation doesn’t bring a technical improvement, it may not meet the patentability threshold. - Overlooking copyright or data risks.
Many AI systems train on copyright protected data. If your data isn’t properly sourced or licensed, you could face legal issues, especially as laws around AI ethics develop.
Where AI is driving innovation (and patents)
At Stratagem, we’ve helped protect AI-led inventions across several sectors:
- Healthcare and bioinformatics – from diagnostic tools to clinical decision systems.
- Automotive – self-driving technology and intelligent traffic management.
- Smart manufacturing – process optimisation and predictive maintenance.
Thinking of patenting your AI solution? Start here.
If you’re considering patent protection for your AI-enabled invention, it’s essential to begin with a clear and strategic plan. Filing too late or without the proper preparation can lead to missed opportunities or, worse, no protection at all.
Here’s a guide to help you get started on the right track:
- Identify a technical effect
Not all AI applications are patentable. You’ll need to show how your invention improves a process, solves a technical problem, or produces a measurable outcome beyond basic data handling. The presence of AI isn’t enough; it’s the impact that counts. - Name a human inventor
AI can’t be listed as an inventor. You must name the person responsible for the inventive step, typically the one who designed, guided or interpreted the AI’s output. Make sure you document this clearly during development. - Avoid early disclosure
Sharing your idea too soon via pitch decks, demos, or online content can make it ineligible for patent protection. Use NDAs for early discussions and consult with an IP advisor before going public. - Review freedom to operate (FTO)
An FTO assessment helps uncover existing rights that might block your path to commercialisation of your innovation or expose you to risk. It can also inform your product roadmap and shape your commercial rollout. - Decide between patents and trade secrets
Not all AI innovations are best protected with a patent. Some algorithms or data models may be better kept confidential. We can help you weigh up whether patenting or secrecy aligns best with your goals and timelines. - Check your data rights
Training data can raise copyright and licensing issues. If your model uses third-party or scraped data, it’s vital to ensure its appropriately licensed. This is especially important when scaling your solution or pursuing investment.
Taking these steps early can make the difference between a strong, enforceable patent and one that misses the mark. More importantly, they ensure your AI and IP strategy supports the long-term success of your business.
Why patenting AI-driven innovation matters to your business strategy
IP isn’t just about protection; it’s a lever for growth. When your invention uses AI, securing a patent can add real value across your business:
- Attracting investors. Patents signal credibility and foresight. They can help demonstrate your tech edge in due diligence.
- Boosting valuation. IP is an intangible asset. An extensive patent portfolio can materially impact how your company is valued.
- Opening doors to partnerships. Tech collaborators and commercial partners often seek protected intellectual property (IP) before committing to a serious engagement.
- Supporting licensing and revenue. A patent gives you control. That means you can license your innovation and generate returns while continuing to innovate.
Protecting your AI solution isn’t just a legal move; it’s a strategic one.
Frequently asked questions
- Can I name an AI system as the inventor on a patent?
No. Most jurisdictions require a human inventor. Even if AI contributed to the idea, a person must be named, typically the individual who guided or interpreted the AI’s output. Documenting that human involvement is essential for your application to succeed. - Can I patent software that utilises artificial intelligence?
Yes, if it delivers a technical effect. AI that enhances a system’s performance or solves a specific problem may qualify. But tools that automate routine tasks without technical impact likely won’t. The presence of AI alone isn’t enough; it’s the outcome that matters. - What if I talk about my AI invention before filing?
Public disclosure before filing can destroy novelty. Sharing your idea in pitch decks or meetings without protection risks making it unpatentable. Minimise pre-filing disclosure and protect any absolutely necessary discussion by having them under NDA early and consult with an IP professional before going public. Filing early is one of the most effective ways to secure your position. - Can I use copyright instead of a patent for AI-generated output?
Sometimes. Copyright may apply if a human meaningfully contributes to or selects the AI’s output. Fully autonomous AI-generated content, without human authorship, is unlikely to be protected under current copyright law in most jurisdictions. - Can I patent a prompt engineering method?
Potentially, yes, if your method delivers a technical effect. For example, if it improves system performance or enables a machine to complete a new task, it may qualify. However, purely creative or generic prompts are unlikely to meet the patent threshold without a clear technical contribution. - What if my AI system evolves after filing?
Your patent protects the invention as originally filed. If your AI system develops new capabilities or changes substantially, you may need to file an additional application. Keeping your IP strategy adaptable ensures your protection evolves with your technology and avoids leaving valuable innovations exposed.
Conclusion: AI is a tool; it is neither an inventor nor inventive in its own right.
AI is a powerful enabler of innovation but doesn’t rewrite IP rules. You still need to demonstrate technical contribution and name a human inventor.
Whether you’re a start-up or tech company using AI in healthcare, automation or infrastructure, don’t assume the presence of AI makes your idea automatically patentable.
Need help navigating AI and patents?
We’re here to help. Our experienced patent attorneys support clients daily in defining, protecting, and commercialising their AI-driven inventions.