Find out how a lack of understanding and accountability in government spending could be burning your tax dollars in the rush to fund AI projects.
Talking about AI is all over the place these days, and debates on risks, ethical concerns, and copyright issues are getting plenty of airtime. These issues are important, sure, but there's another immediate danger when one topic grabs all the headlines: the waste of tax money that could be better spent.
It's a pattern we've seen before: a topic gets hot, and suddenly, tax money is thrown at it – often so hastily that it ends up misplaced or squandered. Take the post-pandemic period, for example. Many countries faced scandals over mask procurement, with officials caught in bribery schemes. Then there's the shady €35 billion vaccine deal between the EU and Pfizer, now under scrutiny by the EU prosecutor's office. The push against global warming has also seen its share of issues, like the questionable practice of carbon offset trading and climate aid disappearing. The U4 Anti-Corruption Resource Centre warns that "corruption within climate finance threatens the global achievement of [climate] goals".
And now, AI is falling into the same trap. As non-expert decision-makers (our representatives) are allocating huge sums to areas they don't fully understand, we need to be vigilant. This article exposes instances where government funds have been wasted on so-called AI initiatives, explore why this keeps happening, and advocate for transparent use of tax money.
The Problem With IT
In 2024, the Biden administration earmarked billions for AI across various departments, including a $1.8 billion grant for the Department of Defense to adopt and deliver AI capabilities. However, the U.S. government’s track record with IT projects is dismal. From 2003 to 2012 only 6.4% of federal IT projects with labor costs of above $10 million were considered successful. The same analysis found that 52% of large projects were "challenged", and 41.4% as straight-out failures. Issues ranged from overambitious project scopes to reliance on outdated systems and complex stakeholder involvement. The takeaway? When government money flows into IT, especially on a large scale, there's a high chance your tax dollars are going down the drain.
Furthermore, AI investments have been misappropriated. The U.S. Department of Housing and Urban Development, for instance, was granted funds for surveillance to curb crime using facial recognition. However, this technology was misused by public housing authorities to spy on residents and harass them over minor housing violations and, in some cases, to evict them based on the surveillance footage. Such occurrences should raise alarms about the efficiency and appropriate use of large-scale IT investments, especially in AI.
Big Money for a ‘Thin Layer on Top’
When government officials jump on a bandwagon, they often don't really understand what they're buying into. A recent incident in Europe is a prime example:
At the start of the year, the Austrian public employment service Arbeitsmarktservice (AMS) unveiled its AI chatbot, Berufsinfomat. AMS introduced it as a digital transformation flagship, costing €300,000. However, it was soon exposed by hacktivist Mario Zechner and others, that the chatbot is essentially a ‘very thin layer on-top’ of ChatGPT, created by a company called goodguys.ai, which sells off-the-shelf software that use OpenAI's API. Furthermore, the frontend code appears to have been largely generated by ChatGPT itself. Given the exorbitant cost of €300,000, the project's justification is highly questionable. Worse, the Berufsinfomat faced backlash for biased responses, suggesting stereotypical career paths based on gender, such as IT or trade to a male, and gender studies or philosophy to a female high-school graduate. When confronted with such criticism, Johannes Kopf, chairman of the board of AMS, responded that “We have already achieved a lot. We're still on it.” – thus implying that they have any meaningful control over the ChatGPT-based bot’s output, a notion that is far from the truth.
This illustrates how the officials in charge of such investments have little idea what they are buying, and what the actual possibilities and limitations of such software are.
What Can Taxpayers Do?
Transparency and accountability are key. The implementation of digital platforms that allow real-time tracking of government expenditures might be the best solution. This would grant us a level of transparency that would make it much harder for those in charge of public spending to waste large sums on AI projects with little substance, or on frivolous expenses like $20,000 trashcans. Government bids must be transparent and open to public scrutiny, ensuring that simple software solutions are not excessively overpriced. Moreover, special attention is needed in scrutinizing AI implementations, given their potential for bias and impact on privacy.
It’s actually quite straightforward: private lives should remain private, and public decisions and expenditures should be transparent. Perfect transparency might not be the preference of decision-makers who are comfortable spending money they didn’t earn, but in 2024, there’s no excuse for taxpayers not to have access to every invoice funded by their taxes, ensuring clear insight into how public money is spent.