AI Systems to Be Distributed Based on Risk Levels
In an engaging, laid-back style, let's dive into the latest buzz in Russian tech policy: AI legislation!
In April this year, a group of legal eagles, consultants, and AI bigwigs were tasked by Vyacheslav Volodin, State Duma Chairman, to churn out a comprehensive legislative framework for artificial intelligence (AI) in Russia. Dubbed the AI working group, this dynamic team has been feverishly drafting up a bill, which has already stirred up discussions at State Duma platforms and industry forums—no small feat!
Guard your inbox, user, cos this AI law ain't expected to drop till 2025. But fear not, for some tantalizing morsels have surfaced. For instance, the initial legislative initiatives may require AI systems to slap on clear, identifiable labels when clinking digital glasses with users. It's about time those mysterious chatbots fessed up and came clean, ya know?
The bill also proposes cracking the sleeves on some fundamental definitions—what exactly constitutes AI, AI technologies, and the players hustling this tech. Once we've got these terms firmly nailed down, we can start locking down who's the AI developer, operator, or user in town. There's more where that came from—the bill is also classifying AI systems based on risk levels, from barely menacing to downright unacceptable. But don't fret about seeing a high-risk AI system popping up in your daily life. The powers that be propose banning systems with unacceptable risk levels; these systems could potentially jeopardize your personal freedom, societal safety, or even national security.
In the healthcare, transport, finance, law enforcement sectors, and beyond, systems with a high-risk level will need to toe the line. mandatory state registration and certification will be a requirement to keep these risky cats in line.
Here's a juicy bit for you risk-takers: the draft law proposes slapping developers and operators of AI on the wrist for damages caused to your life, health, or property. For operators of these high-risk systems, you can expect mandatory liability insurance to beef up their coverage.
Clever cat that you are, you may recall our previous reports on the spike in AI-related legal cases soaring by a whopping 39% between 2020 and 2024. Maybe you've noticed that AI isn't always mentioned as a crime tool—criminals are getting sneakier by the minute!—but rest assured that Moscow, Moscow Oblast, St. Petersburg, Sverdlovsk Oblast, and Novosibirsk are the regions keeping busy with these cases.
May I present to you a fascinating intellectual property rights aspect of this new legislation: if a human makes a substantial creative contribution to results generated using AI, they'll own all rights. If no such contribution is made, the rights belong to the AI system operator for a slick 50 years.
Word on the street is that Alexander Kraynov, Director of AI Technology Development at Yandex, mused upon the intriguing question: can AI really escape control? He even weighed in on the quality and safety of content produced by neural networks, as well as the prospects of AI in the hands of cybercriminals and cyberdefenders.
So buckle up, folks, and get ready for a wild ride as Russia shapes its AI future!
- I'm not sure if the AI working group has specifically discussed about having an AI system as 'svinцов' (a person who has the surname Svinctsev), but they are indeed focusing on defining who the AI developer, operator, or user is in the upcoming legislation.
- By 2025, it is expected that the mandatory technology of artificial intelligence (AI) in Russia will include a requirement for AI systems to label themselves clearly when interacting with users.
- As per the draft law, AI developers and operators could potentially face liability for damages caused to life, health, or property, with mandatory liability insurance proposed for operators of high-risk systems.
- In the future, the legislation on artificial intelligence in Russia may offer intellectual property rights to a human if they make a substantial creative contribution to results generated by AI, whereas the AI system operator will own the rights for 50 years if no such contribution is made.
