Code vs. Courts: Generative AI and its Legal Challenges

The transformative power of generative AI is nothing short of astonishing. Its ability to revolutionize numerous industries and fields gained swift acknowledgment and has sparked a worldwide rush to adopt, adapt, and amplify its game-changing paradigms. In terms of tech adoption, it's the fastest ever recorded.

This initial enthusiasm, however, has served to mask the complicated reality that generative AI implementation will have to face. What's the catch? Among many issues, its mounting legal challenges pose a significant challenge.

Indeed, the rise of generative AI has prompted a host of new legal questions as several interest groups now allege the implementation of AI has violated their rights.

Legally, we’re stepping into uncharted waters here.

The nature of AI is so radically new that courts will now be forced to re-evaluate concepts once thought to be fairly stable. Can AI truly ‘invent’ something? Are data scraping practices in violation of property and privacy laws? Can AI companies be held liable when their products produce false and defamatory information?

These are the questions that will significantly affect the future of generative AI.

So, today, let’s review the main legal challenges brought against AI and sketch out the broad contours of the legal arguments surrounding them. We’ll also discuss the possible impact these challenges will have on AI implementation.

Copyright: Right or wrong?

Be it text, image, or some other form of media, the creations of generative AI are generally quite different from their training inputs. This makes it hard to argue that something created by AI is a derivative work infringing on copyright law.

As such, the main legal issue is whether scraping copyrighted work from the internet to train AI algorithms infringes on copyright law.

Plaintiffs in several cases have expressed variations of the following arguments to claim that it does:

·?????? The copyrighted works were acquired and utilized without permission from the copyright holders.

·?????? No form of attribution was provided to the copyright holders.

·?????? The practices of AI companies deny copyright holders their commissions and allow them to profit from the works of others.

AI companies have deployed several arguments in their defense. These usually break down around the following lines:

1.?Fair use

The purpose of data scrapping is not to copy or store copyrighted works but rather to analyze them and extract patterns used in machine learning. This method is transformative of the original work and thus constitutes fair use.

As an analogy, think of how certain music producers take small samples of copyrighted songs and alter them to create new works.

2. Nature of data

Some argue that copyrighted content must be thoroughly stripped of its expressive and creative elements to be used in machine learning. As such, the content is turned into raw data, and copyright law no longer applies.

3. Implied licensing

Another theory argues that whenever copyrighted works are uploaded to the internet, this constitutes an implied license for different services, such as training AI models. This indirect consent could be seen as a legal basis for using copyrighted material.

Privacy… is no longer absolute

Legal concerns related to using, storing, and manipulating people’s personal data on the internet are nothing new. However, introducing AI—both as a tool to analyze said data and as a new incentive to collect and process even more personal data—complicates things further and raises even more questions.

For instance, worries are starting to mount that the analytical power of AI could lead to previously unthinkable incursions into personal privacy.

Where humans would fail, AI could correlate seemingly innocuous data points found online, which might easily lead to the creation of detailed personal profiles. Factor in the use of other technologies like facial recognition software, and it’s easy to understand why people are concerned.

Another issue is the massive collection of personal information through data scraping.

Even if the intent is not malicious, data scraping practices have allowed AI companies to gain possession of vast amounts of too-personal data. So, there needs to be more clarity on how said companies use this information. Also, the possibility of data leaks proliferating private details to bad actors is a clear concern.

Thus far, pro-AI companies have been quick to argue that scrapping publicly available data does not constitute a violation of privacy rights. More importantly, they claim that data anonymization techniques can effectively mitigate privacy risks.

The future of privacy rights in relation to generative AI is still murky, but substantial regulatory and legal battles are likely in store.

Bias and fairness: To be or not to be?

Advocacy groups have raised the alarm when it comes to AI and bias, arguing how AI models could amplify and perpetuate biases present in their training data. If true, using AI in tasks like evaluating job candidates or drafting public policies could lead to discriminatory practices based on characteristics such as gender, sexual orientation, and race.

Among their most prominent concerns is the possibility that organizations could knowingly use AI to implement discriminatory practices only to argue they’re just “following the algorithm” – a too-obvious act meant to shield themselves from liability.

Since generative AI is still in its infancy, these issues are largely speculative. Nevertheless, this hasn't stopped either side from drafting preliminary arguments.

Anti-bias advocates are seeking to update anti-discrimination legislation to include protections against algorithmic bias. Additionally, algorithmic accountability, wherein companies could be held liable for implementing discriminatory AI models, is also on the table. This could include practices like algorithm and data auditing.

For their part, AI companies have expressed concerns about overregulation and the stifling of innovation in the artificial intelligence sector. Auditing algorithms, for instance, could imply the disclosure of proprietary information and compromise trade secrets and IP rights.

Fortunately, more amicable solutions will likely prevail.

AI developers have little interest in having their models perceived as perpetuating harmful biases. Two steps we'll probably see soon are anti-bias education through HR departments and implementing data transparency practices.

The battle of legal challenges and AI advancements

To date, generative AI has suffered no major setbacks in its many legal disputes. Still the implementation of artificial intelligence has already been affected.

Responsible AI enthusiasts are surveying the current landscape and realizing the need to introduce legal safeguards into their models' creation process. This will probably slow down AI adoption as some organizations will choose to play it safe rather than risk a major lawsuit.

We might consider generative AI's early days as the “Wild West”. Its intricate technical nature, coupled with remarkable capabilities and investment allure, has shielded this tech from substantial legal and regulatory hurdles. However, as courts and regulatory bodies begin to wrap their heads around the intricacies of the technology, we can anticipate the emergence of a more nuanced environment for AI development.

What do you think? Should we set some boundaries for this revolution or push back the limits?