Perhaps no technology has created as much uncertainty, as much ‘superposition’ of possibilities in recent decades, as Artificial Intelligence (AI). Is it a fad or is it here to stay? Will it take jobs or add more? Does it steal intellectual property or not? After several years, 2025 will be seen as the year when answers started to crystallize.

Ethical Trends

For one thing, it is increasingly clear that ethically ambiguous uses of AI are neither speculative, nor few and far between, but are widespread and being heartily embraced. Companies are using AI to hire, even though it is unclear whether this is indeed the best path forward. AI systems are also commonly making inaccurate and biased decisions depending on who is being assessed. For instance, a recent study revealed that generative AI tools recommended more advanced diagnostic tests for high-income individuals compared to lower-income patients, despite identical symptoms.

Another prominent example is the use of AI in academia, both learning and assessments, but also teaching, and evidence is clear that some of these uses are creating a ‘cognitive debt’. Scholarship, journalism, literature and the entire internet is now riddled with AI slop, a term for mass-produced AI-generated content that is often generic, repetitive, and published with minimal human oversight or editorial judgment. The fears that these risks would come to pass had existed for a while, but now the fears have visibly crystallized into reality. The long-term implications of these trends are far from certain, but that they inspire confidence is inconceivable at the moment.

This is not to undermine AI’s positive potential. In education and scientific research, AI has made useful contributions too. Indeed, in 2025, AI delivered systems that detect dementia with over 97% accuracy and helped engineer viruses that kill harmful bacteria.

AI And IPR

Some AI related matters are already in courts. The most litigated matter from the inception of the generative AI models has been that of intellectual property rights, especially copyright. As early as January 2023, such lawsuits had been filed, and over 50 litigations remain in the courts currently.

In 2025, answers emerged, and fair use played a decisive role. A US district court ruled that training AI models using legally acquired copies of the books was not copyright infringement. The court adjudged that the training of the models was clearly ‘transformative’, a factor employed in favor of a fair use finding, and which means that the defendant created something different. In an interesting turn of events, the court denied this defence where pirated copies of the copyrighted material had been acquired from illegal sources. For the latter, the AI company Anthropic wound up having to settle for a sizable sum of 1.5 billion dollars. Adding a balancing element in the jurisprudence, in a different case with similar facts, the court opined that even though there was a transformative use, if there was evidence that complainants’ market had been significantly harmed, fair use protections could be denied. Outside of the US, a UK case decided that models themselves could not be treated as infringing copies.

AI And Jobs

After years of theoretical projections, 2025 research has finally documented measurable employment effects from AI adoption. A report by a team at MIT and Oak Ridge National Laboratory found that AI is already capable of replacing up to 11.7% of the workforce in the US. This is supplemented by another study, which concludes that by late 2025, AI has significantly reduced traditional entry-level roles, shifting the job market toward graduates who possess AI and data skills, or other specialized problem-solving skills.

On a positive note, and in a contrary finding, a June 2025 analysis by PwC of a billion job advertisements found that contrary to fears of displacement, sectors highly exposed to AI are seeing 38% increase in job availability and AI skilled workers are getting a 56% wage premium compared to non-exposed sectors.

Regulation and Geopolitics

Mitigation of the negative impacts of AI, and therefore regulation in some form must remain a priority. Over the past few years, the European Union spearheaded the regulatory shaping of AI through the passing of its AI Act, and it appeared plausible that the world will follow suit. However, while some jurisdictions, like South Korea and Italy made some progress on this path, 2025 has not seen much progress globally, with even the European Union diluting some of its regulations. Much of this reversal can be attributed to the fear of losing the global ‘AI arms race’, increasingly projected as a winner-takes-all scenario. The current geopolitical landscape adds fuel to the fire.

This shift can be seen in the release of manifestos by the two rivals, USA and China, which frame technological dominance not just as an economic goal, but as an existential necessity for geopolitical survival. For years, the United States has tried to militarize the top of the AI stack by blocking China’s access to cutting -edge GPUs that could boost its AGI ambitions, while China has tightened its grip on key raw materials that the United States needs to stay competitive in the wider race for industrial supremacy. 

The geopolitical rivalry is the key to unlocking any global consensus on AI regulation. As long as major powers treat AI as a strategic arms race, nations are effectively trapped in a prisoner’s dilemma in which no government can easily afford to regulate unilaterally.

Uncertain Future

One thing is clear: AI is no fad. It’s not even a standalone phenomenon, like the pandemic. It’s fast becoming an intrinsic part of the socio-economic fabric. And while 2025 saw some clear trends emerging, the path forward remains less than certain. The biggest source of uncertainty is whether the current trajectories will yield exponential improvements in capabilities, or will plateau, requiring fresh thinking. 

On the regulatory end, while the implications of the ongoing arms race will become clearer over the next few years, the end of this race doesn’t appear in sight, barring some divine intervention. 

At the individual level, working with AI efficiently and responsibly, avoiding cognitive debts, would remain the best path forward.

(Krishna Deo Singh Chauhan is Associate Professor, Jindal Global Law School and Associate Director, Cyril Shroff Centre for AI Law and Regulation; Anupriya is Lecturer at the Jindal Global Law School. Views expressed are personal. They can be contacted at agodbole@jgu.edu.in)