The European Union has agreed to pass the world’s first laws governing the use of artificial intelligence. It is one step on a long and winding road.
It is unsurprising that this initiative came out of the EU. It has been the only governing body to consistently put its people ahead of the wishes of the companies that control the search and social media platforms that intrude into virtually every nation on the planet.
The historic agreement came after 36 hours of solid negotiation among the EU member states and it sets out the parameters on which the laws will be based.
The move is hugely significant, but it should not be seen as a full solution to curbing negative impacts while allowing the positive aspects of AI to flourish.
It aims to ensure that AI systems used inside the EU are safe and respect fundamental rights. In other words, it is based on a harm principle. That means it will target high impact AI systems that pose potential risks and strictly limit the use of potential AI tools for state surveillance.
What that is likely to mean in practice is that high impact systems – and that would include systems such as ChatGPT and Google’s new Gemini AI system – will be forced to meet strict transparency obligations and could face fines of up to seven per cent of their global revenue for violations.
EU states will be banned from using AI surveillance such as remote biometric identification except in cases of immediate need such as an imminent threat of terrorist attack or investigation of serious crimes. Police and security forces will not be allowed to routinely employ AI for untargeted scraping of facial images on the internet or CCTV footage, predictive policing, or biometric categorisation and social scoring. The laws will not stop police forces from using AI to catch criminals, but they will stop agencies from using it to predict who might commit a crime.
An AI Office will be set up within the EU to enforce the planned laws.
Private sector high impact systems will be affected not only by the new transparency rules but also by a requirement such systems undergo a “fundamental rights impact assessment”. The agreement states this must be done before a system is introduced to the market but, given the fact that ChatGPT and Gemini have already been deployed, it is likely that there will also be retrospective requirements.
What it does not address is one of the greatest threats to news media: The commercial impact of generative AI, which presently has the ability to plunder intellectual property at will as it creates iterative results to users’ queries. I doubt that scraping news archives will raise a red flag in a fundamental rights impact assessment that is firmly rooted in the rights of individuals. There will be an obligation to disclose that the product of generative AI used sources such as a newspaper article in its ‘creation’, but that is a long way short of requiring the AI engine to seek permission let alone pay for it.
Of course, the vulnerability of intellectual property to exploitation by generative AI is not solely a problem for news media. Such systems exploit all of the creative industries and much, much more.
The European Union has started in the right place. Harm to individuals and the power of the state must take priority. And it has taken the EU more than two years of talking to get to the point where they have an agreed framework for laws. Their work must not end there. Commercial and wider social impacts of IA must also be addressed…and sooner rather than later. If it does not happen, the adverse impacts that search and social platforms have already wrought on society will seem mild by comparison.
One of the champions of the new laws, Romanian member of the European Parliament Dragoş Tudorache, has said that the EU was determined not to make the mistakes of the past, when tech giants such as Facebook were allowed to grow into multi-billion dollar corporations with no obligation to regulate content on their platforms including interference in elections, child sex abuse and hate speech.
If other MEPs share the same view, they must move on the broader social and commercial impacts as well.
New Zealand will see some benefits from the EU laws. Europe is too big a market for the multinationals to turn their back on it to avoid legal obligation. It would come as no surprise, however, to see Meta’s Mark Zuckerberg or X’s erratic Elon Musk imposing short bans to put the frighteners on governments with less resolve than the EU.
For the most part, AI systems will have the proscribed harmful aspects engineered out in order to meet EU regulation and the moves will be embedded across all markets. That cannot be said of commercial activities, where financial opportunities will continue to be exploited in individual markets because they will be easier to tailor.
Last weekend, the Guardian quoted a Columbia Law School Professor Anu Bradford saying the EU initiative could “set a powerful example for many governments considering regulation. “[Other countries] may not copy every provision but will likely emulate many aspects of it”.
Will New Zealand be one of them? I hope so but I will not be holding my breath for immediate action, given that even the flawed Fair Digital News Bargaining Bill has been rumoured to face an early death.
And we should not delude ourselves that laws alone will protect us from the negative impacts of artificial intelligence.
AI has the potential to immeasurably benefit humankind. For example, in Tokyo a couple of weeks ago my son was shown an AI-generated digital replica of an individual’s heart. This exact copy is being used to test drugs and medical procedures. Then possibilities for beneficial use of AI seem endless.
Unfortunately, the negative and sometimes malevolent uses are also numerous.
Transnational digital enterprises are difficult to curb, but it may be impossible to bring to heel states that do not respect the rule of law and criminal elements that sometimes enjoy the protection of such regimes.
Imagine for a moment the damage that could be caused by a picture-perfect avatar of a government minister, spouting dangerous disinformation in a pitch-perfect voice, going viral on social media. Imagine a prominent journalist appearing to do the same thing.
It will be extremely difficult to prevent bad actors from employing AI to perfect deep fakes and fake news that to date have had tell-tale inbuilt flaws.
The only defence against such attacks may be verification and education.
Journalists already verify information whose veracity they doubt, but vigilance may need to become standard practice even on routine matters. How many check back to ask whether a media release had actually been distributed by the organisation to which it is attributed?
The public may have to be equally vigilant and that will mean educating them to access multiple sources and to seek out trustworthy sources that act in the public interest and verify before they publish.
Legislation may have its limits, but it is vital that AI is subject to the rule of law in a way that the Internet was not. When I read of the EU initiative, I was reminded – in a perverse sort of way – of Neil Armstrong’s words when he became the first human to set foot on the Moon: The EU framework will be a small step for Man, but it certainly cannot yet claim to be a giant leap for Mankind.
The right move
Warner Bros-Discovery has made the right move in choosing a single host to fill its 7 pm current affairs slot next year. By using Ryan Bridge as the sole anchor, the show will not be subject to the banter that was part and parcel of The Project, with its four presenters.
It will also be far more cost effective for a network that isn’t exactly rolling in cash. I feel sorry for the three whose contracts will end through no fault of their own. However, they represented a line-up of talent that Three could not afford.
Nor could it continue to shoulder the cost of licensing an Australian format that, frankly, was no longer the epitome of originality and innovation.
The format of the new show has yet to be revealed but there is room for a robust, interview-based current affairs programme as a real alternative to One’s Seven Sharp which, in contrast to its title, is a comfortable-as-old-slippers magazine show.