A bill on AI transparency and responsibility will be submitted in the United States, allowing people to check whether their creations have been used for AI learning
A bill called the Transparency and Accountability for Artificial Intelligence Networks (TRAIN) Act has been introduced in the United States to make it easier to check whether your work has been used without permission for AI training. If this bill is realized, copyright holders will be able to access the learning records and check whether their work has been used.
TRAIN Act Seeks Transparency in AI Use of Copyrighted Works
Senator introduces bill to compel more transparency from AI developers
The TRAIN Act provides that copyright holders can request disclosure of the training records of a generative AI model if they can declare a 'good faith belief' that their work was used to train the model.
Under the bill, AI developers would only need to disclose enough material to identify whether copyrighted work was used, otherwise the law would presume that they used copyrighted work.
'The mechanism is simple: If a copyright owner's work is used to train an AI, they need a way to determine whether it was used in that capacity. And if it turns out that it was used, they should be compensated,' said Senator Peter Welch, a Democrat who introduced the TRAIN Act.
In recent years, the explosive growth of generative AI has highlighted the conflict between technological innovation in AI and the copyright holders who feed on this innovation. Creators in a variety of fields, including images, text, music, and video, are questioning the merits of training generative AI with their copyrighted works.
SAG-AFTRA, the actors' labor union, and the world's largest publishers sign an open letter saying that creative works should not be used to train AI without permission - GIGAZINE
In addition, many of the companies developing generative AI do not disclose their training methods, meaning that creators cannot tell whether their work has been trained by the AI. The TRAIN method is expected to directly address this 'black box problem.'
On the other hand, many large-scale AI models randomly collect large amounts of information, and it has been pointed out that even the developers cannot accurately grasp what data is being collected, and there is a possibility that the developers will not be able to respond to requests for disclosure.
ChatGPT's 'OpenAI' that may have been trained without user consent faces legal challenges in the EU, which emphasizes strict privacy laws; experts say 'compliance with regulations will be nearly impossible' - GIGAZINE
The TRAIN Act has received support from organizations such as the Recording Industry Association of America, the American Federation of Composers, Authors and Publishers, and the American Federation of Musicians, and is praised for its potential to balance innovation with respect for creators' rights.
Mitch Glazier of the Recording Industry Association of America said in a statement, 'Senator Welch's carefully crafted bill will bring transparency to AI and ensure that artists and rights holders have a fair chance to go to court when their work has been trained without their consent. The Recording Industry Association of America applauds Senator Welch's leadership and urges the Senate to sign this bill into law.'
'The future of America's vibrant creative economy depends on laws that protect the rights of human creators,' said Elizabeth Matthews of the American Federation of Composers, Authors and Publishers. 'By requiring transparency about when and how copyrighted works are used to train generative AI models, the TRAIN Act opens the door to creators being fairly compensated. On behalf of our more than one million songwriters, composers and music publishers, we applaud Senator Welch's leadership.'
Related Posts:
in Software, Posted by log1p_kr