On Thursday, Meta revealed its intention to implement more sophisticated AI systems for content moderation, intending to decrease its dependence on external vendors. These AI systems will be tasked with identifying and removing illicit content, such as terrorism-related material, child exploitation, drug sales, fraud, and scams.
The company stated that these enhanced AI systems would be rolled out across its various applications once they consistently surpass the effectiveness of current content enforcement strategies. Concurrently, Meta aims to lessen its reliance on third-party vendors for content moderation tasks.
In a blog post, Meta clarified, “Although human content reviewers will remain, these AI systems are better equipped to manage tasks suitable for technology, including the repetitive assessment of graphic material or fields where malicious actors frequently alter their strategies, such as illegal drug trafficking or fraudulent schemes.”
Meta anticipates that these AI systems will enhance the detection of violations with improved precision, bolster scam prevention, facilitate faster responses to emerging situations, and minimize instances of excessive enforcement.
The company reported encouraging initial results from AI system trials, demonstrating their ability to identify twice the volume of infringing adult sexual solicitation content compared to human review teams, along with a reduction in error rates exceeding 60%. Furthermore, these systems are said to be effective in identifying and preventing a greater number of impersonation accounts targeting celebrities and prominent figures, and in thwarting account takeovers by recognizing suspicious activities like logins from unfamiliar locations, password modifications, or profile alterations.
Moreover, Meta stated that the systems are capable of detecting and counteracting approximately 5,000 daily scam attempts where fraudsters endeavor to trick users into divulging their login credentials.
In the blog post, Meta affirmed, “Specialists will be responsible for the design, training, oversight, and assessment of our AI systems, evaluating their performance and handling the most intricate and impactful decisions.” It added, “Humans will, for instance, retain a crucial role in making the most sensitive and critical judgments, including appeals against account disablement or reporting incidents to law enforcement.”
This initiative aligns with Meta’s recent relaxation of content moderation policies over the last year, a period that coincided with Donald Trump’s return to the presidency. Last year, Meta discontinued its external fact-checking program, opting instead for a Community Notes model reminiscent of X. Additionally, the company eased restrictions on “topics that are part of mainstream discourse” and promoted a “personalized” approach to political content for its users.
Furthermore, this development occurs as Meta, alongside other major technology companies, is currently embroiled in numerous lawsuits seeking to hold social media platforms responsible for adverse effects on children and young users.
On Thursday, Meta also unveiled the launch of a new Meta AI support assistant, offering users around-the-clock assistance. This assistant is being progressively introduced worldwide within the Facebook and Instagram applications for both iOS and Android platforms, as well as via the Help Centers on Facebook and Instagram for desktop users.