The European Parliament has officially voted in favor of the AI Act, and in keeping with the longstanding industry support behind the measure, the IFPI and others are applauding the development.
EU lawmakers today approved the AI Act with 523 affirmative votes, compared to 46 votes against the voluminous legislation and 49 abstentions. Introduced years back, the roughly 90,000-word law was modified several times en route to passage and, as its name suggests, will attempt to address all manner of AI considerations.
The latter include but certainly aren’t limited to the use of AI to compile “facial recognition databases,” the instances wherein law enforcement can legally deploy real-time AI tools, what constitutes “high-risk” AI usages, mandatory labels for AI media, and a whole lot else.
Of particular interest to the industry are the training-related safeguards at hand. As most know, the data and (protected) media upon which large language models (LLMs) are “trained” remains the subject of debate and litigation.
Troublingly, some AI giants continue to express the belief that utilizing an abundance of copyrighted works to develop an LLM, which then draws from and possibly reproduces the information when responding to queries and performing tasks, constitutes fair use.
But the EU’s AI Act explores training-data disclosures, the rights of creators to opt out of training, and much more. (Last year, OpenAI CEO Sam Altman indicated that his company could potentially exit the European Union over artificial intelligence regulations.)
“Any use of copyright protected content [to train AIs] requires the authorisation of the rightsholder concerned unless relevant copyright exceptions and limitations apply,” the AI Act reads in part, also calling on “providers of such models [to] draw up and make publicly available a sufficiently detailed summary of the content used for training.”
Notwithstanding today’s passage, the AI Act must still receive “a final lawyer-linguist check” and formal sign-off from the European Council. Furthermore, the law will only “enter into force” 20 days following its publication in the EU’s official journal, becoming “fully applicable” 24 months after that.
Independent of this full-scale implementation date, though, prohibited practices (like the previously highlighted surveillance and facial-recognition uses) will be barred under the law six months after its entry into force, with the AI Office-developed “codes of practice,” covering “obligations for providers of general-purpose AI models,” set to go into effect nine months after the entry into force.
Meanwhile, “general-purpose AI rules including governance” have a 12-month deadline, compared to 36 months for “obligations for high-risk systems.”
Building on these timetables, the aforementioned International Federation of the Phonographic Industry (IFPI), along with GESAC, ICMP, IMPALA, and many others, is touting the AI Act’s passage – and zeroing in on its implementation specifics.
“We welcome the approval of the EU AI Act by the European Parliament,” the organizations said in part, “and we thank Members of the European Parliament for the essential role they have played in supporting creators and rightsholders throughout the legislative process.
“While these obligations provide a first step for rightsholders to enforce their rights,” they proceeded, “we call on the European Parliament to continue to support the development of responsible and sustainable AI by ensuring that these important rules are put into practice in a meaningful and effective way, aligned with the objectives of the regulation.”
>>> Read full article>>>
Copyright for syndicated content belongs to the linked Source : DigitalMusicNews – https://www.digitalmusicnews.com/2024/03/13/ai-act-music-industry-response/