Meta has unveiled a suite of artificial intelligence tools aimed at improving user support and strengthening content enforcement across its platforms, including Facebook and Instagram.
The announcement, made on Friday 20 March, marks a significant expansion of the company’s use of AI to deliver faster assistance, tackle scams more effectively and reduce errors in content moderation.
Meta AI support assistant launches globally
The centrepiece of the update is the Meta AI support assistant, first previewed in December, which is now being rolled out globally across Facebook and Instagram on iOS, Android and desktop Help Centres.
Designed to provide round-the-clock assistance, the tool aims to resolve user issues from start to finish rather than simply offering guidance.
“When you have an account issue, you need a solution — not just a suggestion,” Meta said.
The assistant can answer queries on topics such as notification settings and new features, while also taking direct action on certain requests within Facebook, and eventually Instagram. These include reporting scams or impersonation, managing privacy settings, resetting passwords, updating profiles and guiding users through content appeals.
Meta said the tool typically responds in under five seconds, significantly reducing wait times compared with traditional support channels. According to early feedback, a majority of users report a positive experience.
The company added that the assistant is available in all languages supported by Facebook and Instagram and will continue to evolve as adoption grows.
AI systems target scams and harmful content
Alongside the support assistant, Meta is expanding the use of advanced AI systems to improve how it detects and removes harmful content.
The company said these systems build on earlier efforts to prioritise serious violations such as terrorism, child exploitation, fraud and scams, while reducing over-enforcement.
Early testing suggests notable gains. Meta said the technology can identify and mitigate around 5,000 scam attempts per day that had previously gone undetected, while also cutting reports of impersonation involving high-profile individuals by more than 80%.
The systems are also said to detect twice as much violating adult solicitation content as human review teams, while reducing mistakes by over 60%. In one example, AI flagged a potential account takeover by identifying unusual login behaviour and profile changes that might appear harmless in isolation.
In another case, it detected a fraudulent website posing as a sporting goods retailer by analysing suspicious pricing and web address patterns, contributing to a 7% drop in views of scam-related adverts.
Meta noted that these tools can operate in languages spoken by 98% of internet users, a significant increase from previous coverage of around 80 languages, and can adapt quickly to cultural nuances, slang and evolving online behaviours.
Human oversight remains central
Looking ahead, Meta said it plans to expand the deployment of these AI systems over the coming years, provided they consistently outperform existing moderation methods.
This shift will reduce reliance on third-party moderation vendors, while increasing investment in internal systems and teams. However, the company stressed that human oversight will remain essential.
“People will remain at the centre of our approach,” Meta said, adding that AI is intended to enhance — not replace — human judgement.
Experts will continue to design, monitor and evaluate the systems, particularly in complex or high-risk cases such as appeals against account bans or reports to law enforcement.
Meta added that its Community Standards will not change as a result of the new technology, and that it is building safeguards to minimise bias and ensure consistent enforcement.
The company said the combination of AI scale and human expertise would ultimately deliver a more effective and reliable system for both user support and platform safety.
No comments:
Post a Comment