Say No to AI Protect Act
Spoiler: One of the promises from the Artificial Intelligence Action Summit held in Paris from November 6 to 11, the European Union is about to vote the AI Protect Act to face the AI crisis.
For some years now, we’ve noticed that the generative AI sector struggles to find its business model. Despite massive investments, it loses money every year. The cause: the efficiency of humans competing with AI models. Behind the buzz, some experts in the field admit, off the record, that the crisis is at a point where jobs are at risk.
To avoid such catastrophe, the AI Protect Act proposes a set of systemic measures and synergy rechanneling by dealing with the heart of the problem: humans consulting content generated by other humans.
Artificial Intelligence Agency
The project was initially intended as a firm prohibition on publishing content generated without artificial assistance (the term replacing "human" in the regulation), and was initially denounced by the Pirate Party because it attacks, in their view, the fundamental right of freedom of speech.
Despite the Greens' support in parliament, who oppose any attempt at artificialization in the name of biodiversity preservation, it was ultimately the Republicans who brought a solution to the institutional deadlock by proposing a fee, part of which would be invested in sustainable data center research.
The current project, which should be voted on without substantial changes (according to sources close to the case), proposes the creation of a new European Artificial Intelligence Agency. Having investigative powers and administrative decision-making authority, it will ultimately be responsible for collecting and redistributing the fees. Until its formal creation, collection will first be organized and managed by national agencies; ARCOM for France.
The legislation also provides that the fee will be due and split between producers and consumers of content generated without assistance. The allocation will be set by decree of the Agency (and individually by national agencies in the meantime), and dissuasive fines are planned for humans passing themselves off as AIs.
ARCOM has already announced the creation of the "French AI Generated" label for content-generating companies (advertising agencies, publishing houses, the press). After an independent audit and review of the case by the Agency, these companies will benefit from a simplified reporting regime and their listing in a national directory allowing citizens to easily find assisted information sources.
Self-Hosting
The entire indieweb sphere has risen up against this measure «taken by Brussels bureaucrats despite the advice of technical experts». We know how difficult it is to determine the human proportion in their website statistics, but the stakes are high since it will serve as the basis for the fee they must now pay.
Cloudflare and Google have enthusiastically announced the modification of their respective captchas to fit the new context. It's been known for a long time that it's increasingly difficult to find tasks that only a human can complete, for AIs are increasingly good at passing themselves off as humans.
Both companies should soon provide a new inverted captcha allowing any webmaster to identify robots by asking them to solve problems too difficult for humans (e.g., finding Charlie).
This arms race between detectors and humans promises to be exciting, as a Moldovan hacker who gained access to these captchas has written a Firefox module that detects them and uses AI to solve them. The experts we've consulted draw a parallel with age verification for pornographic websites and predict that these technical solutions cannot work with hosting providers.
Internet Service Providers
As a complement to fee collection, Internet Service Providers would have to do their part and have divergent views. Some shout about the attack on Net Neutrality (a convenient scapegoat), while others see innovative opportunities for their subscribers.
According to our sources, Bouygues and SFR should integrate content logging to enable detailed end-of-month billing. Free is considering a plan where subscribers get it all included.
For its part, Orange, in partnership with ARCOM, will offer an update of its Hadopi-labeled filtering tool, available since 2009, to include a whitelist of French AI Generated-labeled sites, thus offering its subscribers secure, free browsing.
Beyond providers, the entire tech sector is buzzing with excitement, and we've heard of a joint project between Microsoft and the Ministry of Culture to natively offer similar filtering functionality in Windows 11 and Edge, allowing citizens to be exempt from the fee.
Online Petition
Even if the text should remain largely unchanged until the parliamentary vote, the reactionary forces of the web still hope to reverse the trend. La Quadrature du Net and Framasoft have launched an online petition calling for the outright withdrawal of the project.
To learn more: I can only advise you to check the article's date.
Social Media
The big tech companies have announced their willingness to respect the new regulations and to work on innovative new features for their users.
LinkedIn has already announced that human users with a standard account will only see and be seen by AIs and other robots, and, for the nostalgic, the creation of a Premium Authentic subscription that will not have this limitation.
Following the merger of xAI and X (formerly Twitter), Elon Musk has announced that all content published by free users will now be reformulated by generative AI to "exonerate themselves from European woke imperialism". Premium users, meanwhile, will be able to choose from several reformulations.
Diametrically opposed, Mark Zuckerberg believes AI is meant to assist users in their daily lives by replacing the friends’ content feed with an AI-assisted synthesis of network news tailored to your preferences.
Off the record, a Meta executive admitted: "there's been no human content for a long time."