Nov 13 2026
Could Outdated Online Information Become the Next Major AI Liability Risk?
A summary of Lawrence Shaw's article for Emerging Risks, and what it means for brokers and their insureds
TL;DR
AI systems repeat what they find in an organization's published information, including material that is outdated, inaccurate or contradictory. This summary of Lawrence Shaw's Emerging Risks article sets out where liability questions could arise, why D&O underwriters are paying attention, how insurers are changing their wordings, and how brokers can arrange an independent assessment of an insured's misinformation exposure.
On November 13, 2026, Emerging Risks published an article by Lawrence Shaw, CEO of AI visibility assessment company AAAnow, on a source of AI liability that has received less attention than hallucinations and model accuracy: the accuracy of the information AI systems read and use to create their answers. This page summarizes the article and sets out what it means for brokers.
The full article is available to read on Emerging Risks.
As published on Emerging Risks
Voices section, November 13, 2026


From model accuracy to the information AI reads
Most of the focus on AI misinformation to date has been on hallucinations, model accuracy and the governance of AI systems. The article identifies a further source of potential liability: the information AI systems have access to. Businesses hold large quantities of it across websites, including third-party sites, intranets, document repositories and databases, and some of it may be inaccurate, outdated, contradictory or poorly controlled. AI has made that information easier to discover, combine and redistribute.
An AI system could give incorrect financial information about a business, outdated safety instructions or policy wordings, or old guidance on how to take a medication. If someone acts on that information and suffers a loss, the liability questions that follow could be complicated.
When poor information becomes a liability
The article gives two anonymized examples from different sectors.
Two anonymized examples
From the Emerging Risks article, organizations in two sectors
Pharmaceutical
What was published
Information about a banned drug remained publicly accessible, alongside outdated guidance on its administration.
What AI could do with it
Retrieve and repeat potentially dangerous information that could be reused by a third party.
Heavy industrial machinery
What was published
Inaccurate servicing and safety instructions were introduced into an AI system supporting engineers.
What AI could do with it
Critical isolation instructions were not given at the appropriate point in the servicing process.
Neither example establishes an insured loss or a successful liability claim. Both show how weaknesses in information management could carry more consequence once AI makes existing material available for operational decisions.
The article lists the possible implications for businesses:
- Professional negligence
- Misrepresentation
- Product liability
- Regulatory scrutiny
- Directors and officers
An emerging D&O consideration
The insurance industry already treats AI misinformation as a concern. In Gallagher’s 2026 AI Adoption and Risk Survey of more than 1,200 businesses, AI errors, misinformation and hallucinations ranked as the top AI-related risk.
Top perceived threats from AI adoption
Percentage of businesses citing each threat, 1,200+ businesses surveyed
Bar chart of the percentage of businesses citing each threat, from Gallagher's 2026 AI Adoption and Risk Benchmarking survey of more than 1,200 businesses. AI errors, misinformation and hallucinations: 57%. Legal and reputational risks from AI misuse: 56%. Data protection and privacy violations: 55%.
| Perceived threat | Businesses citing |
|---|---|
| AI errors, misinformation and hallucinations | 57% |
| Legal and reputational risks from AI misuse | 56% |
| Data protection and privacy violations | 55% |
Source: Gallagher 2026 survey, AI Adoption and Risk Benchmarking.
Claims experience and underwriting interest
A 2026 survey of 250 insurance professionals by The D&O Diary and Allianz Commercial found that 85.6% agreed AI-related regulatory, shareholder and operational claims will significantly affect D&O claims experience.
Aon and Lockton have highlighted underwriting interest in AI governance, board oversight and the accuracy of corporate disclosures. Allianz Commercial has flagged AI hallucinations and AI washing as emerging D&O exposures.
AI-related claims and D&O
Insurance professionals who agreed AI-related claims will significantly affect D&O claims experience
Donut chart. In a 2026 survey of 250 insurance professionals by The D&O Diary and Allianz Commercial, 85.6% agreed that AI-related regulatory, shareholder and operational claims will significantly affect D&O claims experience, and 14.4% did not agree.
| Response | Share of insurance professionals |
|---|---|
| Agreed | 85.6% |
| Did not agree | 14.4% |
Source: The D&O Diary and Allianz Commercial, 2026 survey of 250 insurance professionals.
There is an important distinction between an AI system producing an incorrect answer and that answer originating in information the organization itself was responsible for maintaining.
Lawrence Shaw, Emerging Risks
For directors, the question could become whether reasonable steps were taken to identify and address known weaknesses in corporate information. A failure to do so would not automatically establish liability; that would depend on the relevant legal duties, management’s responsibilities, evidence of the alleged failure and the resulting harm. Information governance could still become a more relevant consideration when board oversight is assessed, as the use of AI to generate and search for information grows.
When the AI belongs to someone else
Businesses do not necessarily control the AI systems using their information. Independent platforms draw on public websites, company documents and other published sources, and could retrieve an obsolete financial report, an outdated product specification or a superseded corporate statement and present it as current. Publication of outdated information does not by itself establish liability for an external platform’s answer. Each link in the chain would have to be established.
What would have to be connected
From the published source to a loss, as set out in the article
Source
Information the organization published, such as an obsolete report or superseded statement.
AI output
An external AI service retrieves it and presents its contents as current.
Reliance
Someone relies on the answer and acts on it.
Loss
A loss results from that reliance.
With AI-driven research now common, organizations may find their publicly accessible information under closer scrutiny.
How insurers are responding
For insurers, the cause of misinformation could become as important as the technology involved. The article notes the following changes.
| Insurer | What has changed |
|---|---|
| Munich Re | aiSure underwriting examines AI inputs, training data and outputs. |
| AXA XL | Gen AI cyber endorsement covers data poisoning of training data, not only the client's own data. |
| Beazley | Explicit AI wording added to its cyber and technology E&O policies. |
| Hiscox | Explicit AI wording added to its technology professional indemnity cover. |
As AI becomes more integrated, insurers may want to understand how organizations produce authoritative information, manage document versions, correct inaccuracies and prevent obsolete material from being relied on.
Support for brokers
For an insured, the practical challenge is how to obtain an independent audit and an outside view of its exposure to misinformation. AAAnow supports brokers in raising this with their clients without the broker having to take on the subject.
How a request works
Three steps, from the broker to the insured
Broker requests
The broker names the insured and selects D&O liability, company exposure, or both.
AAAnow assesses
What AI can find and interpret on the insured's website and in its PDFs.
Insured receives the summary
Where its own web pages or PDFs could contribute to a misleading answer.
What the summary does
Describes the insured’s exposure to likely misinformation.
What it does not do
Establish liability for the company or its directors and officers, or determine cover.
If the insured commissions corrective work and monitoring, it can document that work as evidence of mitigation. Brokers can also arrange a call about working together. Requests are made through the AAAnow broker request form.
