After two days of intense questioning, the Centre on Friday shifted its discussions with Meta to technical issues, with the company outlining measures to tackle deepfakes, child sexual abuse material (CSAM), bot accounts and unlabelled AI-generated content on its platforms, sources said.
Sources said that the government has asked Meta to implement specific measures and report back on the outcomes, adding that officials will regularly monitor the company's progress. Another round of discussions is expected next week.
The technical meeting focused on how Meta intends to address the government's concerns, including the continued circulation and virality of harmful AI-generated content even after it has been flagged.
Officials also questioned why AI-generated videos without labels continue to remain visible on Meta's platforms despite the Information Technology Rules requiring synthetic content to be identified and labelled, sources said.
The government also pressed the company to strengthen human oversight of content moderation, particularly by improving understanding of Indian languages and local nuances.
According to sources, Meta explained how its moderation systems work and outlined proposed solutions to address deepfakes, CSAM and bot accounts. The company said it has mechanisms to tackle these issues and detailed how they could be deployed.
Government sources stressed that the objective was not censorship but ensuring compliance with Indian laws.
Sources had earlier said the company admitted there were "serious issues" and assured it will take continuous action to resolve them.
"The government will continue to press the company to undertake sustained measures, actions and improvements to address its various concerns," the sources added.
The discussions follow two days of meetings during which IT Minister Ashwini Vaishnaw and senior officials questioned Meta's global team, led by global affairs chief Joel Kaplan, on its recommendation algorithms, safeguards and compliance with Indian laws.
The meetings were convened after Meta temporarily restricted Prime Minister Narendra Modi's Facebook post.
On Thursday, officials questioned the company on whether it was adhering to Indian laws, how its algorithms function, and the guardrails in place to ensure compliance with constitutional and legal requirements.
Meta had assured the government that it is serious and "working hard" to address concerns relating to deepfakes, child sexual abuse material, bots and synthetic content.
On Wednesday, Meta apologised for the temporary removal of Modi's post, as well as for failures ranging from child sexual abuse content to deepfakes and boosting certain content by being paid "a lot of money", government sources had earlier said.
While Kaplan said in a written statement on Wednesday that he "apologised to the minister on behalf of Meta for the error restricting PM Modi's post", government sources, speaking on condition of anonymity, had said the apology came from company chief executive Mark Zuckerberg.
A Parliamentary Standing Committee has also sought an apology from Zuckerberg over the removal of Modi's Facebook post, warning that the platform's legal protections could be reviewed if no apology was received within three days.
Modi had first posted the selfie video on Instagram before sharing it on Facebook, where it was temporarily restricted. Meta later restored the post, attributing the removal to an error by AI-powered automated content filters and apologised.
According to government sources, Vaishnaw also asked Meta on Wednesday whether the company accepted payments to promote content that could trigger public disorder.
Officials also cited Section 79 of the Information Technology Act, telling Meta that failure to meet intermediary obligations could jeopardise its Safe Harbour protection.
Last month, the government served a notice to Meta over alleged child sexual abuse material in paid advertisements on Instagram. The meetings this week have focused on issues including CSAM, deepfakes, safeguards for verified and prominent accounts, algorithmic bias and compliance with Indian laws.
(With inputs from PTI)