Microsoft’s head of AI declares all online content up for grabs

A recent interview with Microsoft’s head of AI Mustafa Suleyman offered a surprisingly candid insight into the global legal framework. What he said, in a nutshell, is that all content freely available on the World Wide Web can be copied, reproduced and edited in any way.

In an interview with CNBC, Suleyman, who presents himself as having advocated for AI safety and even ethics for over 15 years, shared his views on many fundamental questions pertaining to artificial intelligence. 

Ill-gotten gains?
One of the issues that was raised is that the large AI models have been trained using material protected by copyright (i.e. ‘intellectual property’) without the corresponding usage rights having been granted by copyright holders or other owners of the rights. Suleyman was therefore asked whether AI companies can be said to have stolen IP.

The Microsoft expert’s answer was that the prevalent ‘social contract of that content since the 90s has been that it is fair use – anyone can copy it, recreate with it, reproduce with it’. 

Suleyman then goes on to mention a separate category covering cases where websites, publishers or news organisations have explicitly only permitted scraping for indexing purposes – a ‘grey area’, in his words, which the courts will have to settle.

So while the New York Times is suing OpenAI and Microsoft for using countless publications by the renowned media outlet to train ChatGPT, the Microsoft AI expert fancies himself back in the 90s. 

The global legal status quo
To make matters worse, the international dimension of the issue and the pertaining rights are being disregarded, as OpenAI’s solution is available across the world and Microsoft is a key global player.

Even if practices were to be considered as ‘fair use’ in the US, this does not mean in any way that they comply with European law. Ethically speaking, referring to business practices such as those of OpenAI as a legal ‘grey area’, to quote Suleyman, is also rather questionable, to say the least. The basic tenet of copyright law is that the author holds all rights over their creative work. Only they are entitled to exploit their work in a tangible form.

While some questions regarding copyright and AI still need to be addressed, the fact remains that training data is reproduced both during the pre-training and the fine-tuning stages. As this does not only occur temporarily, this particular use case requires permission. In the event that the holder of the rights has not consented to their work being used, statutory exemptions are the only way to go. In the case of AI training, the reproduction of works may be permitted if this is required for their automated analysis, provided that the rights holder has not explicitly reserved this type of exploitation. However, this is conditional upon the material having been accessed lawfully and being deleted after the end of training. The rights holder can also object to this use by means of reservation (also known as an ‘opt-out’).

Who cares about legislation?
Nevertheless, the statement by Microsoft’s AI expert does not offer such a nuanced view. He sweepingly announced that all public content can be used freely, which reverses the principle of copyright and of rules and exceptions, extending the American idea of ‘fair use’ to all publicly accessible content. The current situation is also deemed to be a ‘grey area’ in his own words, and there are obviously no qualms about the ethical alignment with the company’s corporate values.

If that’s the case, then should that not apply to Microsoft content too? This would include online computer imaging software such as Windows, Office and Teams, among others. It would hardly be in Microsoft’s interest if it were possible to indiscriminately copy and edit their content, yet the AI expert gives no reason why the same rules shouldn’t apply to them.

Another factor is that it is not just one single user consuming content on the free web but OpenAI, a company into which Microsoft has invested multiple billions of dollars and that has made a global business out of using content generated by third parties without their consent.

It is therefore not surprising under such circumstances that business interests prevail and that the law of the jungle has triumphed. There is simply no place for compliance with copyright law or even ethical principles. Any legal ambiguity is ignored and the rules are unilaterally bent to suit their own ends. It is then up to the affected parties to take legal action – a lengthy and (especially in the US) eye-wateringly expensive path. This seems to be how Suleyman pictures the future. In the meantime, everyone just carries on and it becomes the new normal.

Under the influence of Microsoft 
Microsoft’s AI expert’s opinion and OpenAI’s customs in this field are typical of the software giant’s practices in all other respects. 

The European Commission has even launched investigations against Microsoft due to its distribution of Teams as part of the Office pack, which, according to preliminary findings, equates to a dominant market position. Microsoft’s President Brad Smith seems unfazed.

The charge from the EU antitrust authority will not automatically directly lead to a resolution of the issue, according to Smith. To him, official orders from the EU’s highest competition body only serve as a starting point for negotiations. Due to its market power, the American behemoth seems to dispose of its own international legal framework that takes precedence over regional sovereignty and can be amended at its discretion and to suit its own interests. 

This should come as a wake-up call for customers. The positions held by Microsoft’s president and the AI expert complement each other perfectly. The American giant will stop neither at abusing its dominant position nor at exploiting editorial content for its own business purposes, to then invoke beliefs held by society in the 1990s to their own advantage through the words of their AI expert. 

My take on this: Microsoft customers should protect their IP
It cannot be assumed that the security of customers’ data is of utmost priority for the software company, as shown by a series of security breaches affecting Microsoft Cloud, which were not even communicated properly. Microsoft seems to care little about the intellectual property of third parties, as long as a given practice is not explicitly defined as illegal. OpenAI’s behaviour and its legitimisation by Microsoft demonstrate that they are using their market position and technology to declare legal and free of charge any practice that serves their own interests, thereby disregarding legal clarity, ethical principles or even decency and fairness.

Customers should therefore avoid the combination of AI and for-profit operators that have become known for such views and practices. Is there any guarantee that everything entered into chatbots or even the Microsoft Recall feature will refrain from encroaching on confidential content and protected IP? And if any issue were to come to light, it could simply be presented as yet another security failing. However, once the data has been exploited, the damage is irreversible. Any user granting access to their data should be absolutely clear about the provider’s intentions. 

Andreas E. Thyen, a trained economist, has not tired of voicing his concerns: ‘For many decades, the dependency on Microsoft has been at an unsustainable level. We should therefore be very alarmed by the statements from Microsoft’s circles regarding the possibility to negotiate on European antitrust legislation and especially the protection afforded by copyright. It is clear from their assertions that they are not serious about complying with applicable laws and that they lack the sensitivity to handle creative works appropriately. By flicking the “AI switch” instead of turning back to European fundamental values such as the right to resell perpetual licenses freely, customers risk clearing the path to the ultimate stage of dependency.’