AI Regulations: Investor questions on recent developments

研报英文原文证据摘录

AI Regulations: Investor questions on recent developments

Sustainability ● Global

4 August 2026

1. Are there any areas of limited regulatory certainty that

present a significant risk for the tech sector?

Yes. We would like to flag copyright and model training in the US.

There has been a lack of legal clarity on whether copyrighted data can be used freely to train AI;

Data scraping brings

this opens up numerous questions around copyright infringement and associated operational,

potential copyright

reputational and financial risks, with US firms facing the greatest challenges in the courtroom,infringement

especially from the publishing and music industries (see figure 1). For example, recently in July,

a USD1.5bn settlement in a class action lawsuit against Anthropic was approved by a federal

judge in San Francisco1 – approximately USD3,000 per work for roughly 500,000 copyrighted

books used to train Claude2. Microsoft states that to resolve the intellectual property (IP)-

related claims, including “copyright infringement and other claims arising from AI training and

output”, the company may need to “stop selling or redesign affected products, or pay damages”;

the firm has “paid significant amounts to settle claims related to the use of technology and IP

rights" and “may continue to do so, which could adversely affect the results of operations”.3

Meta states that these lawsuits may bring adverse outcomes, including changes or cessation of

some or all of its operations, substantial payments to other parties, or licensing on unfavourable

terms, which “may significantly increase its operating costs and expenses”, and accordingly its

“business, financial condition, or results of operations could be materially and adversely affected”4.

Figure 1.