Meta allowed ads for child sexual abuse material to be displayed to users in India, and local authorities have called on the social media giant to explain how its ad checking systems failed to detect these illegal promotions before they went live.
As reported by BBC, Meta apparently let several variations of promotions for child abuse material show up in its apps in recent weeks.
As per BBC: “The ads, seen by the BBC World Service, use terms including ‘rape video’ and ‘child video’ and link users to channels on the messaging app Telegram, where they can buy the material for as little as 99 rupees [around $1].”
BBC said that it reported the offending promotions to Meta, and was initially told that the ads did not violate its guidelines. Meta removed the promotions shortly thereafter.
The BBC report comes as Meta continues to push ahead with plans to replace most of its human moderation staff with artificial intelligence agents. These automated systems are empowered to assess and take action against potentially violating content across its platforms.
Meta did not say whether the offending promotions appeared as a result of its AI process, but it did tell BBC that "no system is perfect, and our review process may not detect all policy violations.”
The incident represents another blow for Meta’s expanded AI push, which the company is hoping will eventually provide a significant additional monetization pathway through AI automation as a service for third-party clients.
Meta has already cut around 20% of its labor force this year, with a significant portion of those roles going to AI agents instead. The company is seemingly using its own operations as an example of how its AI tools can cut down on labor cost.
However, initial problems with its AI systems could end up becoming major impediments to its development.
Last month, more than 20,000 Instagram accounts were breached after a group of users discovered they could convince Meta’s account support agent to grant them access without having to provide any evidence that they were the actual account holders.
Meta said it has since patched the issue, but the scale and scope of how people pose queries to AI chatbots seems to lend itself to an almost infinite range of this kind of exploit.
Meta is also facing challenges in AI development. Reuters reported last week that Meta CEO Mark Zuckerberg told employees that the development of its AI agents is not progressing as fast as they had anticipated, adding that recent staff cuts have not yet delivered the planned benefits. The Information reported that Meta has also implemented restrictions on AI token usage, as its costs skyrocket due to its mandated reliance on AI tools.
These costs are likely why Meta is now exploring new ways to make money from its massive AI investment. That includes the development of a cloud infrastructure offering that could maximize the usage of its expanding data center network.
That increased data capacity had initially been earmarked for Meta’s own AI usage, but now, it seems that the company is reassessing its overall AI push, and even the true value of its possible AI offerings.Issues with its internal use of AI could further compound concerns.
Meanwhile, Meta is facing other PR concerns on the AI front.
Last week, Wired reported that Meta was using an army of contractors posing as minors online to test how competitor chatbots responded to prompts involving suicide, sex, eating disorders and other high-risk subjects.
Meta said this was standard research in order to better understand safeguards and mitigation measures. But perceptually, it’s another knock against its AI push, which is already facing significant public backlash.
If Meta’s AI systems fail to protect kids, that could be the last straw in terms of its approach. That could also have big implications for the future of its business, especially if Meta fails to build more advanced AI models via its much-lauded Superintelligence group.
