India demands beta AIs secure government permission before going public
(2024/03/04)
- Reference: 1709533571
- News link: https://www.theregister.co.uk/2024/03/04/india_ai_permission/
- Source link:
India's Ministry of Electronics and Information Technology (MeitY) issued an advisory last Friday stipulating AI technology still in development acquire government permission prior to release to the Indian public.
"The use of under-testing/unreliable Artificial Intelligence model(s) /LLM/Generative AI, software(s) or algorithm(s) and its availability to the users on Indian internet must be done so with explicit permission of the Government of India and be deployed only after appropriately labelling the possible and inherent fallibility or unreliability of the output generated," [1]states [PDF] the ministry's notice.
The document also outlines plans to use a "consent popup" mechanism to inform users of potential defects or errors produced by AI, and to inform users by labelling deepfakes with permanent unique metadata or other identifiers.
[2]
It also orders that all intermediaries or platforms must ensure any LLM or other AI model product does not permit bias, discrimination, or threaten the integrity of the electoral process.
[3]Indian tech minister vows to stop Google removing local apps from Play Store
[4]Law secretly drafted by ChatGPT makes it onto the books
[5]More and more LLMs in biz products, but who'll take responsibility for their output?
[6]X protests forced suspension of accounts on orders of India's government
Compliance is requested within 15 days of the advisory's issuance. Some reports have [7]suggested that after complying and applying for permission to release a product, a developer could be asked to perform a demo for government officials or undergo a stress test – a process IT minister Rajeev Chandrasekhar said would provide more "rigor."
While the document issued late Friday is not legally binding – for now – it does signal the government's expectations and the future direction of regulation.
[8]
[9]
"We are doing it as an advisory today asking you (the AI platforms) to comply with it," Chandrasekhar reportedly [10]explained , adding that at some point this stance will be encoded in legislation.
"Generative AI or AI platforms available on the internet will have to take full responsibility for what the platform does, and cannot escape the accountability by saying that their platform is under testing," continued the minister, according to [11]local media .
[12]
India has suffered a number of challenges and controversies regarding AI technologies. Google [13]fell afoul in late February when its AI, Gemini, responded positively to a query asking if prime minister Narendra Modi is a fascist.
In fairness to Gemini, it responded that Modi had been "accused of implementing policies some experts have characterized as fascist," and indeed that is the case. But lots of politicians, of lots of political stripes, get called "fascist" by their opponents and the commentariat – it doesn't make it true.
Nonetheless Chandrasekhar subsequently [14]called out Google for violating sections of the country's IT Act.
[15]
Google [16]reportedly responded that generative AI is only a tool and "may not always be reliable." The Chocolate Factory promised it's been constantly working to improve.
Last November India's attention turned to the impact caused by deepfakes when an AI image of actor Rashmika Mandanna [17]went viral .
Once again, an advisory was issued – that one stating that social media platforms need to remove deepfakes within 36 hours after they're reported. ®
Get our [18]Tech Resources
[1] https://regmedia.co.uk/2024/03/04/meity_ai_advisory_1_march.pdf
[2] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=2&c=2ZeWp1mW47fMNOW@9pnTxvwAAABU&t=ct%3Dns%26unitnum%3D2%26raptor%3Dcondor%26pos%3Dtop%26test%3D0
[3] https://www.theregister.com/2024/03/03/asia_pacific_tech_news_roundup/
[4] https://www.theregister.com/2023/12/02/chatgpt_law_brazil/
[5] https://www.theregister.com/2023/09/28/llms_business_risks/
[6] https://www.theregister.com/2024/02/23/x_india_farmers_compulsory_takedown/
[7] https://www.livemint.com/news/india/intermediaries-must-seek-government-approval-to-provide-genai-services-it-ministry-11709373353793.html
[8] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=4&c=44ZeWp1mW47fMNOW@9pnTxvwAAABU&t=ct%3Dns%26unitnum%3D4%26raptor%3Dfalcon%26pos%3Dmid%26test%3D0
[9] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=3&c=33ZeWp1mW47fMNOW@9pnTxvwAAABU&t=ct%3Dns%26unitnum%3D3%26raptor%3Deagle%26pos%3Dmid%26test%3D0
[10] https://economictimes.indiatimes.com/tech/technology/govt-directs-social-media-generative-ai-platforms-to-comply-with-it-rules/articleshow/108162287.cms
[11] https://www.business-standard.com/india-news/meity-asks-intermediaries-ai-platforms-to-follow-it-rules-issues-advisory-124030200492_1.html
[12] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=4&c=44ZeWp1mW47fMNOW@9pnTxvwAAABU&t=ct%3Dns%26unitnum%3D4%26raptor%3Dfalcon%26pos%3Dmid%26test%3D0
[13] https://twitter.com/greatbong/status/1762010292001935598
[14] https://twitter.com/Rajeev_GoI/status/1760910808773710038
[15] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=3&c=33ZeWp1mW47fMNOW@9pnTxvwAAABU&t=ct%3Dns%26unitnum%3D3%26raptor%3Deagle%26pos%3Dmid%26test%3D0
[16] https://www.deccanherald.com/technology/gemini-not-always-reliable-in-responding-to-prompts-google-after-chatbots-response-on-pm-modi-2908597
[17] https://www.theregister.com/2023/11/09/india_deepfake_takedown/
[18] https://whitepapers.theregister.com/
"The use of under-testing/unreliable Artificial Intelligence model(s) /LLM/Generative AI, software(s) or algorithm(s) and its availability to the users on Indian internet must be done so with explicit permission of the Government of India and be deployed only after appropriately labelling the possible and inherent fallibility or unreliability of the output generated," [1]states [PDF] the ministry's notice.
The document also outlines plans to use a "consent popup" mechanism to inform users of potential defects or errors produced by AI, and to inform users by labelling deepfakes with permanent unique metadata or other identifiers.
[2]
It also orders that all intermediaries or platforms must ensure any LLM or other AI model product does not permit bias, discrimination, or threaten the integrity of the electoral process.
[3]Indian tech minister vows to stop Google removing local apps from Play Store
[4]Law secretly drafted by ChatGPT makes it onto the books
[5]More and more LLMs in biz products, but who'll take responsibility for their output?
[6]X protests forced suspension of accounts on orders of India's government
Compliance is requested within 15 days of the advisory's issuance. Some reports have [7]suggested that after complying and applying for permission to release a product, a developer could be asked to perform a demo for government officials or undergo a stress test – a process IT minister Rajeev Chandrasekhar said would provide more "rigor."
While the document issued late Friday is not legally binding – for now – it does signal the government's expectations and the future direction of regulation.
[8]
[9]
"We are doing it as an advisory today asking you (the AI platforms) to comply with it," Chandrasekhar reportedly [10]explained , adding that at some point this stance will be encoded in legislation.
"Generative AI or AI platforms available on the internet will have to take full responsibility for what the platform does, and cannot escape the accountability by saying that their platform is under testing," continued the minister, according to [11]local media .
[12]
India has suffered a number of challenges and controversies regarding AI technologies. Google [13]fell afoul in late February when its AI, Gemini, responded positively to a query asking if prime minister Narendra Modi is a fascist.
In fairness to Gemini, it responded that Modi had been "accused of implementing policies some experts have characterized as fascist," and indeed that is the case. But lots of politicians, of lots of political stripes, get called "fascist" by their opponents and the commentariat – it doesn't make it true.
Nonetheless Chandrasekhar subsequently [14]called out Google for violating sections of the country's IT Act.
[15]
Google [16]reportedly responded that generative AI is only a tool and "may not always be reliable." The Chocolate Factory promised it's been constantly working to improve.
Last November India's attention turned to the impact caused by deepfakes when an AI image of actor Rashmika Mandanna [17]went viral .
Once again, an advisory was issued – that one stating that social media platforms need to remove deepfakes within 36 hours after they're reported. ®
Get our [18]Tech Resources
[1] https://regmedia.co.uk/2024/03/04/meity_ai_advisory_1_march.pdf
[2] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=2&c=2ZeWp1mW47fMNOW@9pnTxvwAAABU&t=ct%3Dns%26unitnum%3D2%26raptor%3Dcondor%26pos%3Dtop%26test%3D0
[3] https://www.theregister.com/2024/03/03/asia_pacific_tech_news_roundup/
[4] https://www.theregister.com/2023/12/02/chatgpt_law_brazil/
[5] https://www.theregister.com/2023/09/28/llms_business_risks/
[6] https://www.theregister.com/2024/02/23/x_india_farmers_compulsory_takedown/
[7] https://www.livemint.com/news/india/intermediaries-must-seek-government-approval-to-provide-genai-services-it-ministry-11709373353793.html
[8] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=4&c=44ZeWp1mW47fMNOW@9pnTxvwAAABU&t=ct%3Dns%26unitnum%3D4%26raptor%3Dfalcon%26pos%3Dmid%26test%3D0
[9] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=3&c=33ZeWp1mW47fMNOW@9pnTxvwAAABU&t=ct%3Dns%26unitnum%3D3%26raptor%3Deagle%26pos%3Dmid%26test%3D0
[10] https://economictimes.indiatimes.com/tech/technology/govt-directs-social-media-generative-ai-platforms-to-comply-with-it-rules/articleshow/108162287.cms
[11] https://www.business-standard.com/india-news/meity-asks-intermediaries-ai-platforms-to-follow-it-rules-issues-advisory-124030200492_1.html
[12] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=4&c=44ZeWp1mW47fMNOW@9pnTxvwAAABU&t=ct%3Dns%26unitnum%3D4%26raptor%3Dfalcon%26pos%3Dmid%26test%3D0
[13] https://twitter.com/greatbong/status/1762010292001935598
[14] https://twitter.com/Rajeev_GoI/status/1760910808773710038
[15] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=3&c=33ZeWp1mW47fMNOW@9pnTxvwAAABU&t=ct%3Dns%26unitnum%3D3%26raptor%3Deagle%26pos%3Dmid%26test%3D0
[16] https://www.deccanherald.com/technology/gemini-not-always-reliable-in-responding-to-prompts-google-after-chatbots-response-on-pm-modi-2908597
[17] https://www.theregister.com/2023/11/09/india_deepfake_takedown/
[18] https://whitepapers.theregister.com/
Re: "labelling deepfakes with permanent unique metadata or other identifiers"
Anonymous Coward
In most cases, deepfakes of living people will already be illegal under various applicable laws, so I'm guessing anybody creating them isn't going to be too fussed about complying with new laws anyway. Whether it's a watermark or more subtle forms of steganography, these would be relatively simple to remove - especially if you configure a relatively simple AI service to remove it.
I work for a regulator, non-compliant stuff is an endless problem for us, and despite having laws and standards that say what is permitted, plus the power to seize dodgy stuff, we play a continuous game of whack-a-mole against companies who choose not to comply with the rules. As they're invariably outside the UK there's limits to what we can do to track back to the source, and the big tech firms remove content or listings when the authorities complain, but do nothing to stop new ones being added. The EU I believe has similar thoughts to the Indian authorities, I daresay the Anglophone countries will go in a similar direction.
We could lock things down (products, software, services, information) to a point where non-compliance is absolutely minimal, and attempts to evade the regulations more likely to be detected than not, but that would be a very different world to the one most Western citizens live in. I foresee AI regulators will be created, and do things to prove their worth, without being fast enough or comprehensive enough to actually deal with the issues they were created for.
"labelling deepfakes with permanent unique metadata or other identifiers"
I really wonder how that will work reliably. I imagine that some kind of watermark will be visible on the video, likely not in the middle. So, if it is at the bottom or along a side, anyone who cares can just crop the watermark out and republish the video without it. If the watermark is on the top, that would likely risk cropping the person's head and that would rather severely impact the efficiency of the deepfake, so maybe the watermark should be on the top.
But, given that I have no idea what the identifiers are supposed to be, I might just be spouting nonsense.