Where Will AI Dissent Go in 2026?

From striking screenwriters to poisoned datasets and anti-AI fashion, 2026 may be the year resistance to AI stops asking nicely

Opinion Rage against the machine is growing, despite a US administration determined to remove all obstacles to winning an AI race with China, including seeking to suppress state-level AI regulation; and in spite of the anticipated flood of IP-related litigation being threatened by judicial indifference.

Let’s take a look at some of the new and maturing engines of anti-AI sentiment and activism, and where a growing mood of dissent might manifest or deepen its hold in 2026.

Labor Resistance

Representing an elite class of labor, actors and writers won early concessions against AI exploitation through historically active unions with a significant flair for PR, and strong celebrity support.

Unions in less glamorous sectors tend to be subject to greater political and intra-body pressures; and the prospect of equivalent strikes and industrial action in these areas is perhaps undermined by the evident willingness of the current US administration to intervene in societal concerns through physical, even military measures.

However, since AI is increasingly perceived as an existential threat to workers, and the current investment fervor is thought by many to be motivated solely by the prospect of cutting or eliminating staffing levels, 2026 may find diverse labor groups and individuals concluding that they have nothing to lose.

Actions of this kind could be suppressed or reduced by more political means – by relying on the same fait accomplis tactics that media mogul Rupert Murdoch used to spring job-killing technology on a generation of print workers in the 1980s.

One comforting point of view, offering a (arguably false) sense of security in the context of this most public of revolutions, is that AI is incapable of delivering on its promise to industry, and that in any case, is only being used as an excuse to fire people who would have been fired anyway. This exploits our tendency to understand the present through the lens of history; but given the unprecedented nature of the AI revolution, and the potential extent of its reach, this may not be a reliable approach.

Another source of succor to the threatened worker is the historical tendency of technology to eventually create new jobs that will replace the jobs lost through innovation. Though the World Economics Forum’s January 2025 report predicts 170 million jobs created vs. 85 million lost to AI in the medium term, twelve months’ interval is a long time both in politics and AI, and many of the events of this first year of a radical new US administration, along with general industry and investment developments over the course of 2025, could temper this optimistic outlook.

Crucially, as ever, we have to consider who has commissioned figures and reports, and to what extent they may be subject to industry or political pressures to craft a favorable view around AI’s effect on society.

But the main plank of reassurance for workers of the world remains the extreme fallibility of AI systems, which are inclined to hallucinate, and which, as they are given greater scope and power, are capable of attracting greater negative headlines.

However, AI has already proven to excel at ‘easier’ tasks and jobs, causing real-world job displacement now, and notably threatening the hiring cultures in tech-related industries. When narrowly-scoped, AI is already capable of taking jobs; and even if it should hit implacable roadblocks for harder tasks further up the employment food-chain, it evidently will not retreat into another AI winter this time round.

Thus there are many ways to reframe the current situation, and to suppress protest around AI in 2026, so that any eventual dissent will be a post facto codicil. It remains to be seen if wider industries will have the foresight and the reach to anticipate and immunize against AI, as Hollywood was able to do (with the exception of the VFX contingent) in 2023/4.

AI Data Centers Under Fire

Even before AI, US state deals for new data center construction often met resistance, not least because these enterprises have very minimal local staffing requirements; their benefit to the local economy is usually not a clear net gain; and attracting their business, in any case, usually involves significant offset of tax burdens, further reducing any apparent benefit.

But the new scramble to build and outfit AI-focused data centers – which has caused a critical scarcity of RAM, and fears about blocking consumer access to electricity (or else massively raising electricity prices) – takes the issue notably beyond NIMBYism or traditional intra-state tax concession wars.

This month over 200 environmental groups urged congress to institute a national moratorium on the building of new datacenters in the US, citing issues not only around inflating costs of electricity, but also climate-related consequences of industrializing AI at current levels of investment.

At the local level, where data center defiance is traditionally most vocal, activist in Michigan have rallied for a statewide moratorium on data center construction:

A protester in Michigan, seeking a state moratorium on data center construction, on December 15th 2025. Source

The potential carbon impact of the sheer volume of data centers anticipated to meet AI needs radically redefines the nature and stakes for data center opposition and grass roots campaigning. In 2026 it seems likely that local and organized national opposition will deepen in the US, and that the force majeure tactics that characterized the US administration in 2025 may be tempered by mid-term caution.

Therefore some kind of sop or placatory move of a more diplomatic kind seems likely to be proffered at the state level next year. As for the wider lobbying from climate groups, recent trends indicate that the current administration can ride out such concerns, and counter the underlying claims behind them. As ever, ‘starting a conversation’ on an already well-established topic remains a practical way to disarm it.

Digital Defiance

Not all dissent requires a protest rally or a picket line, and anti-AI feelings are also manifesting at university labs, software houses, and GitHub, among other places where AI might normally expect a warmer welcome.

The University of Chicago’s Glaze and Nightshade initiatives perform data poisoning, both with the intent of allowing visual artists to make their work effectively ‘untrainable’, and to actively ‘attack’ machine learning systems through poisoning approaches, respectively:

From a YouTube presentation for Glaze, a schema explaining how the system inhibits AI-based generalization. Source

Away from the rather newer accent on diffusion-based rebellion, the older movement against AI-facilitated facial recognition continues to innovate through increasingly-popular adversarial clothing:

Commercially-available clothing designed to foil AI-based recognition systems with ‘wearable attacks’. Source: via https://archive.ph/0af6l

Dutch designer Jip van Leeuwenstein has produced a dystopian anti-AI mask designed to foil recognition systems from all angles, and this older project has gained significant traction on social media channels lately:

The far-from subtle mask was designed years ago, but has resurfaced in various communities as AI has gained prominence in the media. Source

Recently, the founder of the popular Firefox fork Waterfox, responding to the Mozilla Foundation’s new enthusiasm for AI integrations in Firefox, stated that Mozilla was ‘making a fundamental mistake’, and that Waterfox would resist AI. The announcement followed on from a similar statement of intent (‘keep browsing human’) from the Vivaldi browser over the summer.

AI-driven browsers surfaced in 2025 in various incarnations, most notably OpenAI’s ChatGPT Atlas, though the company has conceded that AI browsers may never be secure – a position well-illustrated by the recent incident where Google’s agentic AI Antigravity, a browser with (optional) access to your filesystem/s, deleted a developer’s entire drive through a misinterpretation of the user’s commands.

As a community, Firefox users less enamored of the new AI features are quite vocal on the matter, and a number of them have developed various anti-AI Firefox add-ons, including Block AI, Disable AI, AI Blocker, and AI Content Shield, among others.

As for the prospects for grass-roots resistance against AI in 2026, it seems reasonable to expect that perceived ‘overreach’ of AI into software systems, surveillance culture, and general culture, is likely to inspire corresponding volumes of counter-attacks – and increasing numbers of companies that will decide the anti-AI demographic is potentially more profitable than joining millions of other companies in arbitrarily leveraging AI through FOMO.

Some of the older organizations that rallied early to the anti-AI cause continue to operate, including the Stop Killer Robots coalition, fighting the ‘dehumanization’ related to the rise of AI, and PauseAI, a global political movement coordinated by an NGO seeking to mitigate the diffusion of AI. The community defines itself as ‘tech-lovers’ who came to the ‘sad realization’ that AI represents an existential risk.

Similar movements include StopAI and ControlAI, the latter vaunting 190,000 public signatures on its open statement, and – interestingly – a significant number of British politicians.

An extensive report from May this year from the Humboldt Foundation examined the emergence of anti-AI groups in detail for the first time, and concluded:

‘Our research has revealed that AI – unsurprisingly – represents a major technoeconomic paradigm shift, and has ignited profound, multifaceted resistance anchored in deep-seated socio-economic, ethical, environmental, legal and political thinking and [concerns].

‘This resistance is not an outright rejection of “progress” but represents efforts to shape the future of this technology in a way that is aligned with established human values, including human dignity.’

Conclusion: AI as Asbestos

Anyone who even casually engages with social media is likely to have come across Facebook groups, subReddits or other communities which have banned AI content. Earlier in the current evolution of AI, such groups were usually those directly affected by AI take-up, such as freelance writers and artists; now, however, such warnings are frequently found on more general groups – usually communities affected by the unwanted incursion of ‘AI slop’.

A polarization is coming into view, in this regard, in the form of intolerance for any level of AI-involvement, as evidenced across the updated rules of various online forums – as if generative content was radium, asbestos, or a medical innovation leveraging Nazi research.

The greatest manifestation against AI would, in theory, be to not buy AI products, and to boycott products known to ‘contain AI’. However, as we have seen in part, AI is increasingly being bundled into products and services without consultation, like fluoridation, making the avoidance of AI output practically impossible in many spheres.

The scope for a grass-roots insurgency may be limited, since this particular AI revolution is, above all, a B2B affair. It’s not intended that the market for AI products will rise or fall based on direct consumer demand, but rather that machine learning systems suffuse and saturate infrastructure, so that end-users participate by default.

Therefore the extent to which grass-roots progress and global anti-AI movements can influence the apparently inexorable rise of AI may depend on the extent to which they can influence elections, and the extent to which a sufficient number of business concerns and politicians sense that popular opinion may have turned irrevocably against this new cultural force.

First published Wednesday, December 24, 2025