ARTIFICIAL INTELLIGENCE
Undermining Democracy And Life
Watchdog no. 171 (May 2026)
Suddenly, you are reading everywhere about AI – is it friend or foe? It’s both. Life’s like that, it’s complex stuff. Most of you will have used AI tools like ChatGPT, so you know how useful they can be. But AI, in partnership with social media, is also a tool for social manipulation by the rich. Here’s that story, told relatively briefly in 13 points (unlucky for some!). I’ve started with a less obvious example, because it’s disturbing but also gives me hope people can rise above AI.
AI undermines parenting
AI’s corporate culture of myth, secrecy, lies and public manipulation
AI’s circular economy of valuation-pumping
How Chat GPT works
How good is AI? Very good!
So, what’s wrong with AI?
AI spin vs reality
AI support for plutocracy and autocracy incorporated
AI accelerates global warming
What you’ll pay, and pay, and pay
AI as a tool of global wealth extraction
The future for AI
Controlling AI will require reinventing effective democracy
New Zealander Dr. Miriam McCaleb’s research focuses on parental smartphone use, especially by new mothers. Today’s constant parental phone use (“absent presence”) interferes with the parent-child interactions which are the basis for a baby’s brain development and secure attachment. That’s bad enough but there are a host of apps targeting parents which make phone addiction worse like Huckleberry, Nara Baby, and BabyTime. They claim to track your baby and help you plan, but you’re the one constantly taking your attention away from your child to record its feeds, nappy changes, naps, mood, growth, temperature, etc.
Playing with babies is one of life’s great joys, watching how they follow your silly interactions, wondering what they make of it. But a baby that is ignored is not getting the reliable contact which makes them feel safe, or the interaction which powers their learning. You can’t lie about your phone use though, they’re capturing data on everything you do, so the research findings are very clear.
Miriam found that during the first six to eight weeks after birth, first-time parents in New Zealand are picking up their phone an average 106 to 117 times per day, with some reaching over 300, and an average daily screen time over four hours! Your baby is looking at you, dependent on you for everything – entertainment, learning, visual and language skills – and you’re looking at your phone…
What a surprise though, you don’t read about this in the media because there’s a giant new industry pushing its AI wares. And health professionals don’t dare give young parents guidance in today’s market-driven pro-business personal-choice “culture”, despite the proven negative effects of technoference on parent-infant bonding and child learning. If you’re a parent, search for http://baby.geek.nz/ to read more; I highly recommend Miriam’s charmingly chatty posts for new parents.
I did say I put this example of AI because it gives me hope, and it doesn’t sound so hopeful so far, with parent phone-use setting their own children’s development back. But that learning and bonding lost to children, if we talked about it, would motivate and change parental behaviour. All we need is a revamp of health services so they act on evidence, a publicly funded news infrastructure good enough to displace data-thieving search engines, and compulsory health warnings on phones and apps. No problem, see my last section.
I use OpenAI as an example here because it’s the rising star, and is very well documented in Karen Hao’s epic book “Empire Of AI”. OpenAI was founded on bullshit from the beginning. Funded privately by an alliance of the superrich (Musk, Thiel, Hoffman, Livingston) and techbros (Altman, Brockman, Sutskever), OpenAI lured key researchers away from giants like Google and Microsoft with its non-profit status and promise of sharing the results to benefit society. But in correspondence revealed later, that founding group discussed rolling back their commitments to openness once the narrative had served its purpose, to attract top staff. They were helped by Trump’s rapid rise to the Presidency, which everyone in the industry knew was due to social media manipulation.
Various fantasies abound in the rarified world of AI bosses and staff; effective altruism (the fantasy of get-rich-quick to later save the world with philanthropy); support for a universal income (a sop in response to the lost jobs and increased inequality created by AI); safety (redefined as managing the risk of catastrophic crises initiated by rogue superintelligent computers).
The reality of work at OpenAI was staff trapped between the calculated mythology of Chief Executive Officer (CEO) Sam Altman and the incentives of their own employment. Altman was a sociopath, a tightly focused manipulator who cultivated a friendly image (e.g. sending short, quirky plain-language emails in lower case), and promising all his business partners whatever they wanted to hear, while acting in his own best interests.
He famously made himself the legal owner of the OpenAI Startup Fund through a structure which allowed him deniability. Eventually fired by his own board for continuing dishonesty, he returned as CEO after a laughably manipulative public campaign playing to the self-interests of shareholders and staff as the controversy tanked the value of their equity in OpenAI. Of course they backed Altman.
Like other US techbro companies, OpenAI’s corporate culture was experienced by staff as overwork for the “vision” and unquestioning compliance. While researching “Empire Of AI”, Hao was contacted by insiders who confirmed that their human resources (HR) department had a history of threatening to cancel staff’s equity agreements to pressure them to sign binding non-disparagement agreements, while publicly denying the practice. Hao quotes Altman often and you get a good feel for his style. I was reminded of Joan Baez’s cutting line about Bob Dylan in her song “Diamonds And Dust”: “Now you're telling me you're not nostalgic, then give me another word for it, you who are so good with words, at keeping things vague”.
Think about this line “keeping things vague” next time you read a techbro quoted somewhere; it’s their modus operandi. They have teams of people preparing them to stay on message, meaning to avoid all the real issues like failure to find a sustainable business model, excessive demands for compute time, power and water to grow their company and get rich, data theft, avoiding global governance and taxes. Early on Altman was more careless in public; then, he admired the leadership of Napoleon and compared the creation of OpenAI to building a religion.
On 26 November 2025 the Financial Times stated “OpenAI is a money pit with a Website on top”, based on reviewing HSBC’s modelling of future revenue. Financial analysis of companies like OpenAI has to rely on models and assumptions because OpenAI is privately owned, meaning it doesn’t have to release company finances. Those HSBC guesstimates make some heroic assumptions, like OpenAI achieving a user base of 44% of the world’s population ex China (displacing Google entirely from the search market) and getting 10% of AI users onto paid subscriptions. If those two are achieved, its losses will decline to just $76B per year in 2030. If.
Open AI’s expenditure is also a black hole, but there are known contracts with multiple partners including Microsoft and Amazon to rent the vast new computing power needed to service AI. OpenAI has grown so fast, it doesn’t own much in the way of assets, relying on renting computing capacity, and raising new capital issues to both build new models then run them. Much of that capital raising is circular investment between its partners Nvidia, Microsoft, Oracle, AMD, xAI, Intel, etc. And the result? AI-related companies have grown rapidly to dominate corporate valuations. Will they crash in another crisis? Trend charts like the ones below say yes; this one shows the valuations of AI companies compared to well-managed productive corporations.
People like Altman will be fine if the world economy crashes though. In all of these investment booms they are in from the beginning. A few globally successful tech startups created this tiny group of techbros who have wealth and power beyond nation-states. As the boom goes on new public investment flows in; at the end of the cycle the super-rich insiders know it first with their insider knowledge, and exit first. They’ll be fine; you and I won’t. Each new tech crisis will deliver another round of bailouts, low interest and cheap corporate loans, but for us it will be more high consumer costs, insecurity, unemployment, lowered real wages, and stock losses for retirement funds.
You probably already know that ChatGPT is just one type of AI, described as a “large language model”. Specifically, computers trawl through huge volumes of online content to identify patterns and organise very complex sets of rules and parameters into billions or trillions of lines of code, 90% of it generated by AI. This is called the “model”. All those rules dictate which Internet sources will be selected and summarised in the chatbot’s answer to your question. What lies in that hidden code? What we know from the changes to Google searches over time is a) the rules favour engagement i.e. addiction and b) paid content will be prioritised, once the tool has consolidated a global user base.
But you may not know that alongside data crunching to create each new AI system, reinforcement learning from human feedback (RLHF) plays a crucial role in keeping those computer-created decision-trees accurate. To do that, companies like OpenAI developed global online systems directing piecework to casual computer-literate workers in places like the Philippines (categorising roadside images to create rules for driverless cars) and Kenya (identifying pornography). Was OpenAI categorising pornography to keep you safe? No, they were forced to include pornography in the training for their image-creation software because such a large share of Internet images is pornographic.
“Empire Of AI” documents Open AI’s highly exploitative online piecework systems. Piecework rates were initially high to draw people into dependency, then lowered later; workers were tied to their computer to compete for each new chunk of images; in Kenya, lives were ruined from constant watching of pornography; and once the model was trained, the work vanished. There’s more to AI than chatbots; these companies have been designing inhuman workflow systems for themselves. If you think they will deliver corporate AI applications which let you work shorter happier days, think again.
AI is astonishingly good, if you use it wisely. We should make the most of these tools while they’re free, because it won’t last. Here, I’m using chatbots like ChatGPT and DeepSeek as my examples. You have to use these programmes in a chat to get what you want, so review each reply and fine-tune your follow-up questions. AI isn’t a person, it’s a Web-search and document-summary app; it doesn’t know you, so it needs you to provide context. I’m an analyst, but I can’t do Web summaries anywhere near as well, and AI does it in seconds…
And it talks to you like you’re its best mate! – so what else is going on here? How does this work? Here we have to do a bit of intuiting, because AI companies don’t tell you how their models work. This is really a question of economics; if you know what the incentives are, you know how the business works. For example, if there are incentives for corruption, you know you will find it (if anyone looks). In the case of AI, the incentives are extreme. Companies are hugely in debt, with no viable business model. They have to acquire paying customers in huge numbers fast. So, you know addiction is their core design principle (they relabel it “engagement”).
When you think about it like that, it’s obvious that AI builds a profile of you from your first question to your last, and from your online history, then it adjusts its language to match your own style. AI wants to give you the answer you want. This is not a search for the ultimate truth via the Internet; this is a search for paying customers! Here lies one of the core problems. Whatever you seek, AI will help you find it. Whatever business and Government will pay for, they will deliver it.
Here’s one example from my own use, which shows why AI is a) very good, b) unreliable, c) a thief, d) an addiction, and e) very bad. Quite an achievement for a bunch of code! I had to make an application to Family Court for a welfare guardian order. If you’ve ever done this, your heart just sank at the memory. The forms are in legalese, but if you get it wrong the life of someone you love is left in limbo. Most people just pay lawyers – but for a while, we have free AI. This takes ChatGPT or DeepSeek seconds! It would take me days of Website search, copying, concentration and synthesis. AI’s summary is much clearer than mine, and it’s also more nuanced than my brain could hold; it notes the ifs and buts as well as the key points.
But b) it’s unreliable. AI’s programming will happily make assumptions and not tell you, so you can get wrong advice – exactly what you don’t want in a legal process. Here’s an example. I asked for advice on what is expected when the application form asks “Set out the full description of document”. AI claimed “that section is basically asking you to list every document you’re filing with your application”. I questioned this, pointing out there is a separate covering letter which does this. AI came back “You’re absolutely right to question that — thank you for calling it out. You’ve read the form more carefully than my earlier answer did. Let’s reset and get this nailed properly”.
This is why there are hardly any journalists left. Traditional news media barely exists since search apps stole their revenue, and AI wants more of the same. Corporate lobbyists in the US now outnumber journalists by eight to one (“All The Worst Humans”, Elwood). And in the social space, we watch all those reels from online influencers who happily offer their mailing lists and endorsements to the highest bidder.
AI is built to be addictive. That’s their primary design criteria. That friendly language and encouragement is the result of calculated programming. They want you dependent so they can charge you for the service, while gathering and selling your data to earn revenue and cover their huge development costs. They’ll stop at nothing to move from here, where they have huge debts, to there, where you and I pay their profits.
AI is very bad. As I wrote this, Trump ordered the US government to stop using Anthropic because they wouldn’t remove safeguards which prevent AI use for computer-targeting of weapons and spying on citizens. These companies are on the brink of financial collapse. If Anthropic doesn’t give the US government what it wants, another AI company will (and did; OpenAI stepped in the next day). And if Republican political action committees’ (PAC) funds want to use AI to generate millions of micro-targeted messages to voters, including not to vote, they can. That’s the big picture.
Below that, at our human level, remember how AI took direction from my correction above? If I’m anti-vax and remind AI that pharmaceutical companies can’t be trusted, I will be shown lots more anti-vaccine information summarised. AI wants to addict you, not inform you. Right-wing lobby groups and corrupt politicians have a new tool. I was initially surprised when I read that the techbros are actively supporting online scammers.
Meta/Facebook/Instagram has developed an internal “global playbook” to avoid regulatory oversight of fraudulent advertising on its platforms (Reuters/Jeff Horwitz, 1/1/26). But then I thought – of course, they need revenue, scamming is a revenue stream. If the potential for Government surveillance wasn’t enough to scare us about unregulated AI, losing our bank balance should definitely do it. It will be so easy, since they have all our online data…
The industry’s spin on artificial general intelligence (AGI) is a useful fantasy, providing a convenient cover for the reality that each AI application is a hugely complex set of coded rules and expensive, energy-intensive data scanning, specific to its purpose. The developing company knows just what those rules achieve; who they sell to, and for what purpose.
Real AI applications changing our world now include the global personal-data industry’s resale of your life in data to anyone with money and without ethics; workflow systems which increase human exploitation; medical decision-making systems which don’t measure up when evaluated; tracking systems for police and other enforcement agencies; targeting systems for mortgage brokers and credit lenders which entrench race, gender, and class discrimination; and Israel and America’s use of AI for surveillance and drone targeting.
Promising a future where AGI delivers massive prosperity, reduced working hours, solutions to critical global problems (global warming, health) are the grand promises that divert attention from investigating the real consequences of AI. These companies trained their apps by building exploitative global piece-work systems to get cheap human feedback for their models. They will be very happy to build exploitative AI apps for your boss next.
Chatbots are the most widely used AI application, with their carefully programmed friendly language. But the massive cost of computing power to develop them, then answer billions of online questions is an investment in future dependency and monopoly. Those future profits must be extracted from users by subscriptions; from higher costs in the productive economy to pay for targeted advertising; from dictators for social media propaganda, surveillance, and enforcement; and from elitist Governments who require voter manipulation systems to stay in power.
The convergence of these Internet technologies to give power to the rich is my biggest concern. AI’s new version of Internet search gathers your data and focuses you down your preferred rabbit hole; AI further automates targeted social media campaigning; Governments rely on targeted campaigns to win elections so go soft on regulating monopolistic abuses; citizens are isolated by their focus on online noise and disaffected with democratic involvement.
And in corrupt and militarised societies, AI extends much further. Israel’s tools of social control were exposed very clearly in “The Palestine Laboratory1”, which describes how companies used Palestine as a marketing tool in presentations to sell tools of surveillance, social control, and weaponry to anyone who would pay, including both sides of the Sudan civil war.
Goldman Sachs analysis described the sudden new wave of data centres as driving “the kind of electricity growth that hasn’t been seen in a generation”, with the result that utility companies are delaying the retirement of gas and coal plants and the transition to renewable energy in favour of high profits on 24-hour seven-day data centres. Microsoft has restarted Three Mile Island, the nuclear plant in Pennsylvania which had a partial meltdown in the late 1970s. Musk runs 30 unapproved gas turbines to power the training of Grok. By 2030, at just the current pace of growth, data centres are projected to use 8% of US energy. That much new energy, plus more for electric cars, means it must come mostly from natural gas. Climate change will be more extreme.
To get to grips with this, first you need to look at AI’s income streams to understand how it derives revenue. Taking Facebook/X up to 2025 as an example, advertising (display, video, promoted tweets, direct deals) provides 50–75% of revenue. Subscriptions and product revenue provides 5–25% while “data licensing”, apps and enterprise systems provide 5–20%. Globally the US currently provides the largest single share of revenue, roughly 50–60%; Europe about 10%; Asia around 10%; and the balance from the rest of the world.
Now let’s count the ways you’ll pay in the future. First, Chatbots are soooo helpful most of us will end up paying. Second, we’ll reduce our time for human interaction, the alternative way we used to use to find information, so we’ll be more prone to addiction. Third, there’ll be all these apps which are like your friend or your counsellor; they’ll start out free or cheap, but when you need them (and they’ll know, they have all your data), you’ll pay more.
This boom has to crash sometime, so then you’ll pay for the bailouts. And the worst political parties will be in Government more often with the help of tech-social-media manipulation. AI will take away lots of jobs and give nothing back. Business owners and will get more, wages will fall, jobs will be casualised. Taxes will shrink, and so will funding for Government services. AI will make spying on populations infinitely easier; you’ll either lose your job or shut up and take it. Climate change has uncounted human costs. AI targeted weapons make wars cheaper and civilian casualties higher.
Self-serving cross-investment in AI companies may be a circular economy, but for the productive economy AI is a one-way street, both globally and across classes. Leading AI companies are mostly American, and like Google’s regulation-free theft of global advertising revenue and social media’s social influence for cash before it, AI aims to use its market dominance to grab a share of productive revenues around the world.
Profit margins on compute spending for the paying users have risen steadily to reach 68%. Yes, 68%! This is normal business for the techbros. It’s the same at Microsoft, Google, etc. This is the extractive new world created by Trump and Tech Inc. It’s us against the US now. That profit estimate may omit development costs, but if true, future revenues based on a global user base will be huge. Capital investors will take their profits. Data centres will shovel up the cash. Advertisers pay more to these monopolies to reach their customers, prices rise, people are poorer. Elite managers and tech workers subscribe to AI, increase their power and share of productive revenue; jobs, wages and worker power fall while AI workflows will micromanage your working day.
Just in case you missed it in the last Watchdog, this is important so I’ll repeat it. The latest US National Security Strategy released in late 2025 calls for “cultivating resistance” to the European Union and supporting Rightwing parties across Europe. It doesn’t get any clearer. No regulation = maximum US profits. US-owned monopolies will be taking a slice of revenue from the world’s production of goods and services, while increasing inequality within enterprises and undermining democracy.
In the conclusion of her book “Empire Of AI”, Hao emphasises the need for mandated independent evaluations of corporate AI models; openness about training models and technical specification of models; regulatory oversight of AI interventions in critical systems (elections, job automation, financial products, parent and child manipulation, medical advice; and public education on how AI works), to counter the mystification of AI spin merchants. This is all necessary, but incremental and inadequate.
At the personal level, technology is getting harder to use all the time. We all feel this pain, but this is not an accident. Every aspect of your phone and your computer is now designed to serve the tech companies – to take maximum data, to deny consent, to make sure you store data in their cloud, to need AI as your guide to just use your new computer or phone to do what you bought it for. Soon you’ll need to pay to use AI. For public-use chatbots, I expect the end result will be a Google-OpenAI duopoly, with the US moving to somehow exclude China’s relatively energy-efficient DeepSeek.
In the business world, AI will radically increase global inequality. For 40 plus years we’ve seen the rise of shareholder power, backed by ever-more-overpaid boards and senior executives. Management is now about increasing profits, not improving the business or the product. Market control and monopsony is the goal, not competition. For them, AI is the perfect toolset to strip middle management back to the bone - a small highly-paid and compliant elite, above a powerless and underpaid workforce.
And yes, I definitely think AI will crash the US economy. The rich have done this before and they aren’t stupid; most will shift investments early, so pension funds and small investors take the big hit. They control the US government, so the bailouts will again be designed to increase wealth inequality even through a recession. The preferred neoliberal crisis response has been redefined as mainstream economics; Reserve Bank action to lower interest rates across the board. And for us overseas, it’s hard not to follow the US lead on interest rates without attracting excessive investment and the resulting bubble and bust.
Government and Big Tech are on the same side, against us. Big governments like the US, China and Russia rely on tech because their pro-elite anti-worker policies would see them ejected without public manipulation. I already knew Facebook got Trump elected from Kiwi Sarah Wyn-William’s excellent book “Careless People*”. I recently learned from da Empoli’s “The Hour Of The Predator” that Google got Obama elected too, and two weeks later the Democrats halted all AI regulation. Tech support for the Democrats didn’t last though; Trump’s cronyism and corruption offered higher AI profits, with the bonus of blunt pro-US interventions in global markets. * See my review in Watchdog 169, August 2025, https://www.converge.org.nz/watchdog/69/10.html
Smaller governments like New Zealand, Australia and Britain also have Rightwing parties which spend big on social manipulation, while cash-poor “Centre-Left” parties (I wish) have baked in hierarchical cultures and centralised media teams which suppress bottom-up debate about real change. Creating a different future starts with changing political parties whose electoral strategies rely on data and people manipulation.
All of us on the Left are part of the problem too; we’ve been so busy fighting to hold on to Government services, we haven’t talked about how flawed they are, and how to improve them. It’s hard to sell a bad product; things have to change. To give some examples:
Current public sector management relies on small incremental changes. We need something like an Effective Government Commission, charged with reviewing the world’s most effective Governments and charting a path to transform each of our services
Workers need to stop relying on Government and legislation to increase wages - Governments of all stripes aren’t delivering. We need effective unions that support strikes and support their lowest paid members, not the privileged long-stayers.
We need citizens involved in shaping Government, whether that’s the single-purpose Citizens’ Assemblies which Ireland used to reform abortion law, or something new. Political party membership here is embarrassingly low too; parties need to campaign differently to increase membership and participation
MPs’ lives are crap too - no wonder the only people persistent enough to rise to the top are self-centred rather than consultative; we need to revamp Parliament so they get more time in their electorates, more time working in their areas of expertise, less time sitting/shouting in Parliament
A two-tier Government like Australia benefits from separation and negotiation between the macro-policy level and the service-delivery level; either we become a new Australian state or we amalgamate our many inefficient local councils into effective regional units and expand the services they manage so they become good service deliverers again, not bad contract managers
It’s high time we stopped defending the UN and demand it either becomes democratic, or gets out of the way so developing nations create their own platform
Globally, progressive nations need to align with Europe, against the US
That’s just my current favourites - you’ll have your own. Use AI to research and inform your fight for the future, before it’s too late! And on the Left, we need to adapt to the times. Neoliberals have shown the effectiveness of making big changes when in Government; even when they’re out, it’s still hard to go back. Centrist parties need to move Left again, make big changes, and offer a vision of the future which will convince voters. It’s long past time we learnt to use these new online tools to recruit enough members/supporters to be a real force again. No, I don’t mean endless debates with idiots on Facebook/X/Whatever! I mean accelerating the process of reaching sympathetic people, then involving them.
Those who’ve been involved in the Labour Party will know what traditional campaigning looks like; it’s all controlled from the centre. You can’t print your own leaflets, they’re dull and the limited budget means they arrive late; your candidate can’t voice any views other than official policy (Heaven forbid anything took away attention from the glorious leader); community campaigning starts late, has stilted talking-point templates and targets the “undecided”.
New ways to campaign were pioneered by supporters of the UK’s new Labour Left party and the earlier France in Revolt party. One of the most promising is to start campaigning early, and campaign first in areas where support is high. Emails and door-knocking are used to invite possible supporters to submit questions to the candidate, then a link is sent to join an online Q&A; a simple Zoom call, easy to arrange and time-effective. Immediately after, attendees are phoned to encourage them to join supported door-knocking or leafletting in their local community.
This works because a) it’s personalised b) it’s easy to participate c) it moves quickly from contact to involvement d) you’re building your volunteer base early, so you can reach more marginal voters closer to the election. Similarly, online groups make it much easier for the local branches to organise working groups efficiently. No-one turns up to community halls anymore; times have changed, everyone is busy. We need to do everything differently too.
See my review in Watchdog 164, December 2023, https://www.converge.org.nz/watchdog/64/15.html ↩︎
Watchdog 171 -- May 2026
Non-Members:
It takes a lot of work to compile and write the material presented on these pages. If you value the information, please send a donation to the address below to help us continue the work.
Foreign Control Watchdog, PO Box 2258, Christchurch, New Zealand/Aotearoa.
Email