In so many ways, we’re living through a time when what seemed far-fetched until recently is now commonplace. Whether devising contractual models to underpin new business patterns or aligning the legal and regulatory issues that arise when an entirely new range of services comes to market, Tech lawyers’ creativity will have a big part to play in making the far-fetched a reality in 2025.


Two things stand out as key Tech law trends for 2025 – the astonishing progress of AI and what this means for lawyers; and how organisations’ legal and compliance teams are navigating the flood tide of new Tech regulation that reaches its high water mark in 2025 and 2026.


AI scaling laws: the new kid on the block

Moore’s law – that processor density doubles every two years – is being overtaken by a new rule of thumb: scaling laws – that AI model performance doubles every six months. And it’s this new empirical rule that’s leading to ever faster AI models being called out as the accelerant for reshaping the Tech landscape. But it’s not just faster AI models – the race is on for AI model optimisation by making algorithms, compute power and energy consumption greener and more efficient.


With AI compute demand forecast to exceed supply, scaling laws are also behind Big Tech’s investments in nuclear energy – where Google, Amazon and Microsoft each announced deals in autumn 2024 – and data centres – where alternative asset manager Blackstone predicted in mid-2024:


“there will be approximately $1 trillion of capital expenditures in the United States over the next five years to build and facilitate new data centres with another $1 trillion of capital expenditures outside the United States.”


So, we’ll be hearing more in 2025 about scaling AI efficiently as measured by ‘tokens per dollar per Watt’, (where the token is the basic unit of data that an AI model processes, dollar is cost and Watt is energy consumption) and ‘test-time compute’ (giving AI models more time to think).


Yet some are saying that scaling laws are already running out of steam, prompting the Economist to call the current data centre dash:


“the biggest gamble in business history … will investors lose their nerve, or will AI prove its worth as “agentic” systems become more capable and AI-developed drugs emerge?”


AI agents

AI scaling laws facilitate the convergence of three capabilities:

  • new AI-driven user interfaces that support speech, image and video as input and output;
  • memory that retains context; and
  • new reasoning capabilities that complete end-to-end tasks for the user.


In the words of the Microsoft Annual Report 2024:


“this new world is defined by personal, business and organisational AI agents taking action on our behalf and working in concert as a new input to improve efficiency.”


In this context, a striking prediction from research consultancy Gartner is that by 2030, 80% of humans will engage with smart robots on a daily basis, up from less than 10% today.


2025 will see a burgeoning of AI contractual and regulatory issues on Tech lawyers’ desks as a fast growing range of AI services hits the mainstream. As organisations work to balance the benefits and risks of AI services in their businesses, and as AI projects move from proof of concept to pilot to full roll-out, AI Governance – in the shape of impact assessments, transparency documentation, guardrails, policy statements and internal user requirements – will increasingly occupy legal teams’ time in 2025.


EU Tech regulation (again)

With NIS2 (the second network and information security directive) applicable since October 2024 and DORA (digital operational resilience for financial entities) applicable from mid-January 2025, the new world of EU cybersecurity regulation (see Table 1 below) – not inaccurately called ‘the great re-papering’ – has definitively arrived in 2025, with the AI Act to follow hard on its heels in August 2026.


But the transition to these new rules looks set to be challenging in 2025. This is because rules that are cast as directives (NIS2 and critical entities resilience) need to be transposed into national member state law, and as we go into 2025 only a few countries have passed NIS2 into their own law. This is likely to give rise to a period of uncertainty in those countries that are late, quite apart from substantive differences in what each member state’s law will actually say. Delay and differentiation could well lead to further fragmentation (precisely what NIS2 aims to avoid).


In the longer run, with the incoming US administration promising a bonfire of regulation, it remains to be seen how this EU tech rule toolbox will play out globally, geopolitically and in terms of European competitiveness.


Read the full article here: https://kempitlaw.com/insights/tech-law-trends-in-2025-ai-and-tech-regulation-again/