Ctrl+AI+Reg - Top 5 AI Predictions for 2025 (Christmas special)
To wrap up 2024, here are my top 5 predictions for AI in 2025.
As 2024 comes to a close, I'd like to wrap up the year with my top 5 predictions for AI in 2025.
The first 3 relate to AI law and regulation, while the last 2 relate to tech and geopolitical trends.
These are my personal views only, based on my hobby of studying AI policies/regulations across more than 180 countries. See more on my Global AI Regulation Tracker (English version | Chinese version).
Wishing everyone here a lovely break and happy new year :)
Prediction 1
Agentic AI as the new theme (or challenge) for AI regulation around the world. We’re already seeing the new wave of agentic AI capabilities (e.g. Microsoft’s Copilot Agents, Anthropic’s Computer Use model) which is expected to continue in 2025. Agentic AI poses new complexities for lawmakers, being like a crossover between traditional automated decision making (which drives the agent what to do) and GenAI (which drives the agent what to create). Agentic AI will mass stress-test current AI regulatory models across the spectrum (i.e. horizontal 'AI Acts' versus vertical sector-specific laws), with concepts like “AI system” (or equivalent) getting trickier to define in law as agentic AI becomes more embedded into the background of other applications. Agentic AI will also demand more refined and practical measures around explainability (e.g. how to meaningfully explain the numerous instant hidden micro-decisions that may take place in an agentic workflow), an issue which will likely overtake GenAI hallucination as the ‘headline issue’.
Prediction 2
Expect to see widespread reforms across the world in:
‘anti-scraping’ laws (via both data and copyright law lens) to limit or safeguard the mass scraping of public data by developers for commercial development and training of AI.
‘personality rights’ to protect individuals’ facial and voice data from use in deepfakes.
‘high risk AI’ regulation, though not necessarily in the style of the EU AI Act which is actually losing popularity as a model legislation around the world.
While these trends have already started since 2023, 2025 may see even greater legislative activity especially once some of the major GenAI lawsuits in US/UK come to a close (either via judgment or settlement) which will indicate to legislators on what should be the appropriate policy balance between rights holders and AI developers. UK copyright law reform seems promising and could change the game in 2025/26, with potential conceptual/jurisprudential flow-down implications for common law states like Australia, India, Singapore, New Zealand, etc. See my previous Linkedin post on this topic here.
Prediction 3
Privacy and competition regulator activity will continue strong in major AI markets, but privacy issues may start to 'self-resolve' and 'stabilise' as AI models get smaller to be able to run on local devices (i.e. model quantization) where data can be contained locally on-device. There's already a slow market shift from cloud/off-prem back to on-prem, with open-source quantized models (+ an ever-expanding ecosystem of derivative models) challenging traditional cloud-based models. Expect to see more aggressive investments / partnerships from major cloud/AI companies, which in turn fuel competition regulatory scrutiny and enforcement.
Prediction 4
Crypto, blockchain and decentralised finance to probably recapture the headlines (on par with AI) as increasing global uncertainties, inflation pressures and privacy concerns create the ripe environment for decentralised, zero-knowledge proof, automated consent mechanisms and peer-to-peer technologies to demonstrate their value. Expect to see a surge in productive (rather than speculative) blockchain solutions to pressing AI problems (e.g. federated learning for privacy-secure training), with responsive policy and law to follow behind (maybe starting from 2026).
Prediction 5
Increasingly complex geopolitics between ‘developer’ and ‘deployer’ states, with the former controlling supply while the latter controlling demand. Emerging trend of AI sovereignty could evolve into ‘AI nationalism’, with real implications on day-to-day life (e.g. inflated API costs, different geo-availability of AI services across markets, changing T&Cs/contract terms, increasing KYC and other due diligence checks, job opportunities, VC investment flows, nationalistic media narratives, etc). See my previous LinkedIn post on this topic here.
PDF version
Here are my same predictions in a one-pager PDF format.
Disclaimer: This article is based on my personal views, and not representative of any organisation. This article is not intended to provide legal advice or to be a comprehensive guide or reference.
Want more?
Global AI Regulation Tracker (an interactive world map that tracks AI regulations around the world).
Note2Map (a platform to build and launch your own interactive world map tracker).
Global Tech Law News Hub (a news hub that curates the latest tech law news).
AI Governance Library (a library of AI governance templates, policies and frameworks).