Newsletter 16 - Judging a New Technology
Mapping AI's Evolving Legal and Ethical Landscape
Have you ever wondered if anything that you hear and believe in the AI ecosystem is reliable? Think again.
Now there is Mistral, the European GenAI model challenger, ostensibly changing colors and heading into the Big Tech camp from the “Open and Independent Camp”.
This week, according to Reuters, “Microsoft announced it had made a 15-million euro ($16 million) investment in Paris-based Mistral, and would soon make the company's AI models available via its Azure cloud computing platform.”
The article continued, “A Microsoft spokesperson told Reuters on Monday it invested in Mistral without holding a stake. Later, Microsoft clarified to say its investment would convert into equity in the company's next funding round, a common practice among big tech companies investing in AI startups without putting a valuation.”
Immediately outrage erupted in Europe. “Antitrust authorities are already looking at Microsoft's partnership with ChatGPT-maker OpenAI, with the European Commission earlier warning the companies' relationship could be in breach of EU competition rules.” and then they got surprised by this.
Astute observers also noticed a subtle change, apparently Mistral also removed "Committing to open models" from their website.
A few lessons emerge, we know that the Big Proprietary AI Tech ecosystem is pushing for regulatory capture by encouraging and welcoming regulation around AI. They are cornering the market on infrastructure and sequestering talent. All this powder is being applied to an ongoing fight between proprietary Tech Majors and the Open Alliance as it pertains to model efficacy, cost and applicability.
I wrote about all of this in prior essays.
One thing that is also crystallizing is the role of laws, lawsuits and ethical considerations in the evolution of this technology, sometimes running ahead of established policy and regulatory structures. I think that capital is pouring into the “technical box” and inevitably the engineering, model and infrastructure problems will get sorted out and we will rapidly approach optimal states. I also think that too much focus is paid “inside that box” and not enough on forces that are affecting that box from the outside.
This essay is about imagining and organizing the legal and ethical implications of AI into key overarching themes. So that we can generalize and discern lessons from the legal milestones related to the advent of computers, the internet, mobile phones, and smartphones, speculating that might be what AI will have to contend with.
Your mind goes, “what the ..” at this point.
Here is my thought; if the past is prologue, this exercise will help us form expectations defined by the legal and regulatory framework as AI gets broadly deployed in our society. This will allow us to be better prepared as professionals, investors and leaders for what is to come.
**
Nota Bene: To all the legal professionals who are subscribers, please don't take offense at this “I stayed at a Holiday Inn” amateur approach to the nuanced and complicated legal canon. . and to all readers; I am, of course, not qualified to provide legal advice. Do not take this essay as such. This is not a legal analysis, it is merely an analytic approach to constraints that have historically appeared through regulation and litigation with the introduction of new technology that the systems and the economic parameters have to contend with.
**
But first, let's step back.
My last essay, “Welcome to the Machine” proved to be the most effective in terms of both views and engagement amongst all of my work thus far. Thank you for that, your response is my only reward. It was perhaps the intuitive unease that we sense as our perceived world appears to be modified by GenAI. I speculate that it touched upon a primeval uncertainty that everything is not as it appears. As we all know, uncertainty is the root of anxiety and of fear. We are highly motivated by an intolerance of uncertainty.
Fear and anxiety are fertile ground for politics and consequently for Regulation. In Newsletter 11, “AI's Rulebook Emerges: Deciphering Regulatory Realms and Policy Pioneers”, I shared my research on the emerging regulatory and policy frontiers emerging across the world in response to our collective fear and anxiety with AI’s uncertain implication. To help us manage some of the uncertainty, I shared my framework which I use to organize everything I read and hear about AI into a mental model that makes sense for me. Using that framework I then shared my thinking on energy implications of AI in Newsletter 13, “Energizing AI: Navigating the Power Dynamics of Tomorrow's Tech”.
Today’s topic is the next influencing facet in that framework, Ethics and Legality as applied to AI.
Since the dawn of civilization, legal systems have served as the backbone of societal structure, with legal codes tracing back to as early as 2600 BCE. It's fascinating to observe how our societies are intricately woven through the fabric of these laws, guiding citizen behavior and enabling nation-states to uphold order. With the advent of technology, we've witnessed pivotal legal milestones emerge in response, aimed at harnessing societal benefits while mitigating potential harms. Particularly notable is how governments have intervened to preserve market competitiveness whenever the swift uptake of new technologies threatened to disrupt existing equilibriums. The pace at which regulatory frameworks have evolved since the 1970s is nothing short of remarkable, addressing critical areas like Privacy and Data Protection, Intellectual Property Rights, Cybersecurity, Digital Rights, Competition Law and, Liability, and Labor Issues amongst others. Further, the rise of AI and GenAI introduces new ethical dilemmas, including concerns over Bias, Transparency, Consent, Security, and Environmental Impact, challenging us to rethink our approach to Ethical Usage.
Given current matters, my sense is that the prevalent key areas of legal ramifications that AI finds itself contending with at this point are privacy and intellectual property.
Since the 1970s, the landscape of privacy and data protection has seen profound changes, evolving hand-in-hand with the march of computer technology into our everyday lives. The journey from foundational laws such as the US Privacy Act of 1974 and the Electronic Communications Privacy Act of 1986, to more comprehensive measures like the European Union's General Data Protection Regulation (GDPR), illustrates a dynamic response to the growing presence of digital technology in our society. At its core, this evolution has been driven by a fundamental principle of democracy: the protection of individual privacy from excessive government intrusion. Recently we have seen this play out as expected.
First, the Federal Trade Commission's (FTC) settlement with Rite Aid over its use of facial recognition technology highlights emerging concerns about privacy in the age of AI. The prohibition against Rite Aid's use of this technology for five years serves as a cautionary tale for businesses, emphasizing the importance of implementing reasonable data protection measures.
Then there was the case of Parabon NanoLabs’ DNA-Face Modeling used by law enforcement to first generate a potential face from DNA and then to run it through facial recognition software. The use of DNA to create 3D models of suspects’ faces for facial recognition applications raises profound privacy issues, particularly regarding the non-consensual use of biometric and genetic data. This practice has sparked debate over the ethical boundaries of law enforcement techniques, highlighting the tension between solving crimes and protecting individual privacy rights.
That leads us to the controversy surrounding Clearview AI's facial recognition technology, used by law enforcement to identify suspects by matching their photos against a vast database of images scraped from the internet without consent. Clearview AI's method of amassing billions of images from social media and other online sources, to create a comprehensive facial recognition tool, ignited significant privacy concerns. This practice has led to numerous legal challenges, underscoring the conflict between technological advancements in crime-solving and the imperative to safeguard individual privacy rights. The litigation against Clearview AI has spotlighted the critical need for stringent regulations governing the collection and use of biometric data, reflecting the broader societal dilemma of balancing security with personal freedoms in the digital age.
Looking Forward: The principle of "Who watches the watchers?"—originally articulated by the Roman poet Juvenal—remains ever relevant, encapsulating the ongoing debate around oversight and accountability in the digital age. As AI technologies continue to push the boundaries of data collection and processing, ensuring robust privacy protections and maintaining the trust of the public will be paramount. The mentioned cases and regulatory developments signify the initial steps toward addressing these complex challenges, setting the stage for a continued evolution of privacy norms and regulations in the AI era.
Maybe we should have expected this but GenAI and intellectual property rights (IPR) are a flammable mixture which appears to be converging to the mean path of effective use for collective progress. What emerges is a complex picture, one that's been evolving since the early days of personal computing. What possibly started from the seminal Apple vs. Microsoft legal dispute over GUI (Graphical User Interface) elements, laid the groundwork for understanding IP disputes in the digital age. The slew of cases that followed highlighted the challenges in distinguishing between inspiration and infringement, setting precedents that resonate in today’s AI-centric legal quandaries.
First the New York Times vs. OpenAI case underscores the tension between content creators and AI developers over the use of copyrighted material to train AI models. The lawsuit reflects broader concerns about fair use, copyright infringement, and the ethical implications of using vast amounts of data without explicit consent. The settlement and its terms, while specific, are seen as a bellwether for future agreements between AI companies and content providers.
Contemporaneously, there was a licensing agreement between OpenAI and Axel Springer, the German Media company (Owner of Business Insider and Bill Ackman’s nemesis) which represents a proactive approach to addressing copyright concerns. By forming paid partnerships with content creators, AI companies can navigate the legal landscape more effectively, setting a precedent for compensating creators while utilizing their work to train AI models. This model suggests a path forward that respects copyright laws while fostering innovation.
And finally, we all noticed that the agreement between motion picture unions and studios on AI usage in film production. This exemplifies the collaborative effort to define ethical guidelines for AI in creative industries. It highlights the importance of ensuring creators are compensated and credited, addressing potential disruptions AI may cause in traditional content creation processes.
Whereas we are beginning to see some solid legal pathways when it comes to human content creators and the use of their works in training GenAI models, what constitutes new work when GenAI produces it? There is an ongoing argument that says that the GenAI product is no more a derivative than when writers or artists are inspired by consuming previous works. In fact if we are not careful we might end up in situations where AI has more access to such works for derivative creation than humans making the playing field inherently unfair. Ben Sobel is a scholar of information law and a postdoctoral fellow at the Digital Life Initiative at Cornell Tech. He argues passionately in The Economist, “So don’t decelerate copyright. Do the opposite in order to heighten copyright’s contradictions. Show the powerful just how harmful it is when the law stymies learning. Give ai firms a choice: support reforms that eliminate the copyright doctrines that inhibit human and machine learning alike, or watch investments in ai crumble under the same copyright liability that human learners face.”
To drive this point home, in the UK, a landmark ruling stated that AI cannot be listed as an inventor on patent applications. This case, involving applications filed by Stephen Thaler for AI-generated inventions, emphasizes the legal system's current stance on AI and creativity. The ruling signals the challenges AI poses to traditional notions of authorship and invention, necessitating a reevaluation of patent laws as AI's role in innovation grows. Here in the US, U.S. Patent and Trademark Office (USPTO) recently declared that, as with other intellectual property, only a person can receive its official protections. According to Techcrunch, “the USPTO guidance makes it clear that while AI-assisted inventions are not “categorically unpatentable,” AI systems themselves are not individuals and therefore cannot be inventors, legally speaking. Therefore, it follows that at least one human must be named as the inventor of any given claim.”
In recent years, the cybersecurity landscape and the integrity of information systems have become increasingly pressing issues, underscored by notable incidents in the financial sector and critical infrastructures worldwide. This heightened concern has prompted a flurry of actions from governments, policy makers, regulators, and law enforcement across the globe. Notable among these initiatives are seminal pieces of legislation such as the Cybersecurity Act of 2015 in the United States, which aimed to bolster national cybersecurity defenses through improved information sharing between the government and private sector. Similarly, the EU's NIS Directive marked a significant step in enhancing the overall cybersecurity posture within the European Union by mandating a higher level of preparedness and systematic handling of security risks and incidents. Moreover, the case of Microsoft Corp. v. United States in 2016 centered on whether the U.S. government could compel Microsoft to hand over customer emails stored on servers in Ireland, raising questions about digital privacy and the reach of U.S. warrants internationally. The dispute escalated through the courts, with the Second Circuit ruling in favor of Microsoft. However, the case was rendered moot and dropped in 2018 after the U.S. Congress passed the CLOUD Act, clarifying that U.S. law enforcement could access data stored overseas under certain conditions, thereby directly addressing the legal issues at the core of the case. This action brought to the forefront critical questions regarding data privacy, international law, and the extent of jurisdiction over digital data stored across borders. The recent classification of AI as an 'Emerging Risk' by The Financial Regulatory Authority (FINRA) signals a pivotal moment, highlighting the intricate implications of AI on cybersecurity landscapes and the regulatory frameworks for deploying AI technologies in sensitive industries like finance. This evolving scenario underscores the pressing need for a thoughtful approach to harnessing AI advancements, ensuring they enhance security protocols without compromising ethical or operational standards.
Adjacent to the specific laws and litigation, we encounter the quandary of ethical use. We must not only answer the question of whether we can do something but we should further ascertain whether we should. The ethical implications of AI development have burgeoned into a critical concern, especially as these technologies inch closer to the edges of legal boundaries. Ed Newton-Rex, a prominent figure transitioning from AI executive to ethical AI advocate, underscores this pressing issue through his initiative, Fairly Trained. This nonprofit seeks to instill a sense of responsibility in AI companies by certifying that their training data is ethically sourced, much like the fair-trade certifications in other industries. This move not only champions the moral high ground in AI development but also highlights the intricate balance between technological innovation and ethical integrity. The certification demands that data be either licensed, in the public domain, or owned by the company, setting a precedent for transparency and accountability in AI training processes.
Another significant concern is the tainting of AI models with biased data, a challenge that demands rigorous attention to data certification. The polarization in data sourcing practices, as highlighted by a Wired article, reveals a stark divide: while the majority of mainstream news outlets block AI data collection bots to negotiate licensing deals or protect copyrighted content, right-wing media outlets largely welcome them. This open-door policy by some could inadvertently skew AI algorithms towards particular ideological biases, given that AI models reflect the biases present in their training data. This situation underscores the necessity for a balanced and diversified data collection approach to mitigate the risks of embedding systemic biases into AI systems, thereby ensuring that AI technologies serve a broad and inclusive representation of perspectives.
As AI technologies continue to evolve, the need for a comprehensive ethical framework becomes increasingly apparent—one that addresses data sourcing, biases, and the broader impacts of AI on society. The endeavors of individuals like Newton-Rex and the divergent practices in data collection among news outlets highlight the multifaceted nature of these ethical challenges, urging a collaborative effort towards responsible AI development.
These matters, their evolution since the 70s and, the newly emergent legal precedence is sure to establish the guardrails that the AI ecosystem will have to contend with. Overlay on this policy and governance frameworks we talked about previously and you have an idea of what is pragmatically acceptable. All of us as professionals will do well to keep this underlying dynamic in mind as we observe the inexorable ubiquity of AI in our experiences.
Take care of yourself,
-abhi
Sources and References:
Microsoft's deal with Mistral AI faces EU scrutiny By Martin Coulter and Foo Yun Chee, February 27, 2024 - https://www.reuters.com/technology/microsofts-deal-with-mistral-ai-faces-eu-scrutiny-2024-02-27/
Mistral Remove "Committing to open models" from their website | Hacker News - https://news.ycombinator.com/item?id=39517016
From Uncertainty to Anxiety: How Uncertainty Fuels Anxiety in a Process Mediated by Intolerance of Uncertainty - PMC - https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7704173
Cops Used DNA to Predict a Suspect’s Face—and Tried to Run Facial Recognition on It | WIRED - https://www.wired.com/story/parabon-nanolabs-dna-face-models-police-facial-recognition/
This Tech Exec Quit His Job to Fight Generative AI's Original Sin | WIRED - https://www.wired.com/story/ai-executive-ed-newton-rex-turns-crusader-stand-up-for-artists/
Most Top News Sites Block AI Bots. Right-Wing Media Welcomes Them | WIRED - https://www.wired.com/story/most-news-sites-block-ai-bots-right-wing-media-welcomes-them/
Don’t give AI free access to work denied to humans, argues a legal scholar by Ben Sobel - The Economist - https://www.economist.com/by-invitation/2024/02/16/dont-give-ai-free-access-to-work-denied-to-humans-argues-a-legal-scholar
OpenAI Bans Use of AI Tools for Campaigning, Voter Suppression - WSJ - https://www.wsj.com/tech/ai/openai-bans-use-of-ai-tools-for-campaigning-voter-suppression-2308fb98
Eye On AI: More Regulators Look Into AI Investing - https://news.crunchbase.com/ai/regulators-look-into-investing-msft-openai
US Patent Office: AI is all well and good, but only humans can patent things | TechCrunch - https://techcrunch.com/2024/02/12/us-patent-office-ai-is-all-well-and-good-but-only-humans-can-patent-things/