Google has issued an apology to the Indian government over unflattering content on Prime Minister Narendra Modi created by its AI platform Gemini, conceding the platform's unreliability.
Minister of State for IT & Electronics Rajeev Chandrasekhar told TOI that Google had expressed regret over Gemini's unexplained comments on PM Modi.
Chandrasekhar said Google's apology was not enough since labelling the platform unreliable is not a sufficient defence.
He emphasized the importance of AI platforms providing accurate and verified information to users, especially when operating in consumer solutions.
The minister said Google's response to their concerns about Gemini's unreliability was simply "Sorry, the platform is unreliable." He also said AI platforms should not offer consumer solutions while still in a trial phase.
“We had sent them a notice, seeking an explanation on the unsubstantiated results thrown up by Gemini regarding a particular query on PM Modi," Chandrasekhar said.
He said Google's Gemini was an example of impunity, shielded from consequences for potentially violating laws.
The minister's comments came days after he announced that AI platforms in the country would have to seek government's approval before they are deployed for public use.
He claimed India is being used as a testing ground for AI platforms, particularly when they are criticised globally for providing unverified or biased information.
"We have said that the Indian internet is not your lab. If you are moving from the lab and it is still under testing and it is unreliable, you have to put out a disclaimer on the platform saying that this is under testing.
"Also, you have to explicitly inform the user of your platform — in the consent form and through the terms of use — that this is an error-prone platform and this could produce unlawful content… that this is an undertrial platform and may output things that are unlawful and are incorrect," he said.
He reiterated his earlier stance that platforms moving from testing to public use must explicitly inform users about potential errors and unlawful content.
"You cannot consider Indian internet and Indian consumers as an extension of your R&D. You have to respect the Indian consumers, and our digital nagriks. You have to make it very clear to the Indian digital nagriks that your platform may be error prone, may be unreliable, may hallucinate.
"Therefore, tell them in advance that they are dealing with the platform with full awareness and knowledge," Chandrasekhar said.
The minister said that AI platforms could face legal action under Indian IT and criminal laws for breaches and dissemination of inaccurate information.
He warned, "We have the criminal law, and Indian IT Act and IT Rules. These state very clearly that there are 12 types of unlawful content you cannot create. Now if you go ahead and generate unlawful content violative of either the IT Act and rules or the criminal law, you cannot take a defence saying that I'm unreliable. You will be prosecuted."
The minister criticised Gemini for providing "reckless" and "irresponsible" results and showing "disrespect" towards Indian consumers.
On Saturday, the government issued an advisory for AI-led startups, emphasising the importance of labelling unverified information as potentially false and error-prone.
The notice followed another advisory issued approximately two and a half months earlier concerning deepfakes, prompted by multiple instances of synthetic content circulating on social media and internet platforms.
Σχόλια