Professional Indemnity Insurance

Professional Indemnity Insurance for AI and Automation Businesses in South Africa

Professional indemnity insurance covers claims that an AI or automation business’s work, advice or systems caused a client financial loss: a model that produced wrong outputs, an automation that processed records incorrectly, or an implementation that did not deliver what the contract promised. Because insurers are still working out how to treat AI, the policy wording matters more here than almost anywhere else.

Home / Professional Indemnity for AI and Automation Businesses

What professional indemnity covers for AI and automation businesses

Professional indemnity responds to claims by clients for financial loss caused by the AI systems, automations and advice you provide. Depending on the wording, cover typically includes:

Incorrect or misleading outputs

Claims that a model, chatbot or agent produced an output a client relied on and that caused them financial loss.

Automation errors

Workflows or agents that processed transactions, records or communications incorrectly, often at scale.

Failed AI implementations

Projects that did not deliver the accuracy, savings or functionality the client believed the contract promised.

Negligent AI advice

AI strategy, model selection or vendor recommendations that a client relied on and that proved wrong.

Handling of client data

Claims connected to how client data was used to train, fine-tune or prompt a system, subject to the policy wording.

Legal defence costs

The cost of defending covered claims, alongside any damages awarded, within the limit of indemnity.

Is AI covered by a standard professional indemnity policy?

It depends on the wording. Many professional indemnity policies were written before AI became part of everyday client work, so they neither include nor exclude it. Insurers are now adding AI questions to proposal forms, and some are introducing specific exclusions or endorsements.

Check how your professional services are defined. If the policy describes you as a software developer or consultant and you now build AI agents, a claim may be disputed on the basis that the work falls outside that description.
Look for AI and algorithm exclusions. Some wordings exclude or limit cover for losses arising from automated decisions or AI-generated outputs.
Read the intellectual property and data exclusions. Disputes over training data and the ownership of outputs often fall under these.
Disclose your AI work accurately. Not telling insurers what you actually do is a bigger risk to a claim than the AI itself.

Our article on AI use policies and professional indemnity sets out what insurers are asking about, and our page on professional indemnity for technology and software businesses covers the wider software risks.

Where the exposure usually comes from

Most AI and automation claims start in one of these places:

Outcome promises in contracts. Accuracy rates, cost savings or time savings written into proposals and agreements, which turn a disappointing result into a breach.
Third-party models and platforms. Model and platform providers typically limit their own liability in their terms, so when a model gets it wrong, the claim tends to land with the business that delivered the service.
Unverified outputs. Chatbots, summaries and recommendations that reach clients or end users without a human check, where one wrong answer can be repeated many times.
Automated decisions about people. Systems that assess credit, applications, claims or candidates. The Protection of Personal Information Act restricts decisions based solely on automated processing where they have legal consequences or affect people to a substantial degree.
Training data and intellectual property. Whether you had the rights to the data you used, and who owns what the system produces.
Agents that act. Automations that send messages, change records or make payments, where an error is carried out rather than merely suggested.

Where an AI system is also a route into client data, the risk overlaps with cyber. Our page on specialised cyber insurance covers that side.

What Berkley Risk reviews for AI and automation businesses

AI professional indemnity is rated on what your systems do, who relies on them and what you have promised. Before arranging cover, Berkley Risk reviews:

What you build and deploy

Your own models, fine-tuned models or products built on third-party models, and how much of the work is automation, development or advice.

How outputs are used

Whether outputs are advice, decisions or content, and whether a person reviews them before they reach a client or end user.

Contract terms and promises

Accuracy or performance warranties, liability caps and indemnities in client agreements.

Data sources and rights

Where training and client data come from, and the rights you hold to use them.

Testing and human oversight

How models are evaluated, monitored and corrected, and how errors are logged and escalated.

Client sectors

Whether clients are in financial services, healthcare, legal or HR, where AI errors carry higher consequences.

0 +

Years of Experience

0 +

Specialised Products

0 M +

Annual Premiums

Make an Appointment

Building or deploying AI? Fill out the form below and one of our specialists will contact you to discuss your contracts, your use of AI and the cover your business needs.

Professional indemnity for technology and software businesses

Cover for failed implementations, defective code and breach of service levels.

Why SA professional firms need an AI use policy

What insurers are asking about AI use, and how it affects professional indemnity.

Specialised cyber insurance

Breach response, ransomware, business interruption and privacy liability for SA businesses.

Professional indemnity insurance

How professional indemnity works across professions in South Africa.

Frequently asked questions

Does an AI business need professional indemnity insurance?

There is no general legal requirement in South Africa, but corporate and regulated clients often require it in their contracts. It is the cover that responds when a client says your AI system, automation or advice caused them financial loss.

Does professional indemnity cover AI hallucinations?

It can. If an incorrect output formed part of the professional service you provided and caused a client financial loss, the claim is usually framed as a negligent error in that service. Whether it is covered depends on how the policy defines your professional services and whether an AI exclusion applies.

Are AI exclusions common in professional indemnity policies?

They are starting to appear, along with new proposal-form questions about how businesses use AI. Many existing wordings remain silent on AI. Silence is not the same as cover, so it is worth confirming in writing how your insurer treats AI work.

Who is liable if a third-party AI model gets it wrong?

Usually the business that provided the service to the client. Model and platform providers typically limit their own liability in their terms, so the client’s claim tends to be made against you. Your professional indemnity policy responds to your liability, not the model provider’s.

Are claims about automated decisions covered?

Claims that an automated decision caused a client financial loss can fall within professional indemnity, subject to the wording. Regulatory fines and penalties are generally excluded or limited, and the Protection of Personal Information Act restricts decisions about people based solely on automated processing, so these systems need careful design and disclosure.

Does professional indemnity cover AI-related intellectual property claims?

It depends on the wording. Some policies include cover for unintentional infringement of copyright and similar rights, while others exclude intellectual property altogether. Disputes over training data and output ownership are an area insurers are watching closely, so confirm the position before relying on it.

Do AI businesses need cyber insurance as well?

Usually, yes. Professional indemnity responds to client claims about your work. Cyber responds to incidents affecting your own systems and data, such as a breach of a dataset you hold or a system being manipulated to expose information. Many AI businesses need both.

How much professional indemnity cover does an AI business need?

Start with what your client contracts require, then consider the largest loss a client could suffer if your system failed. Automated systems act at scale, so a single error can affect thousands of transactions or decisions before it is noticed. Limits should reflect the volume your systems handle, not only the value of each contract.

Professional indemnity for AI and automation businesses across South Africa

Berkley Risk arranges professional indemnity insurance for AI developers, automation agencies, SaaS providers building AI features and consultancies implementing AI across South Africa. Contact us to discuss your contracts and cover.

Berkley Risk (Pty) Limited (Registration Number 2017/412000/07) is an authorised Financial Services Provider under the Financial Advisory and Intermediary Services Act No 37 of 2002 – FSP#54407

Berkley Risk (Pty) Limited (Registration Number 2017/412000/07)

Authorised Financial Services Provider under the Financial Advisory and Intermediary Services Act No 37 of 2002 – FSP#54407