Today, we’re releasing new AI tools for app support and content enforcement to improve app functionality. As technology advances, we’re applying AI in more ways to give you reliable, action-oriented support when you need it, and to catch more serious violations like fraud faster and more accurately, while reducing over-enforcement mistakes.
Launch of Meta AI Support Assistant on Facebook and Instagram
In December, we previewed the Meta AI Support Assistant. It’s a tool designed to provide reliable 24/7 support for almost any support issue at any time. We’re currently rolling out globally in the Facebook and Instagram apps for iOS and Android, and within our Help Center. facebook and Instagram On the desktop, it offers even more features and helpful methods.
If you have a problem with your account, you need a solution, not just a suggestion. The new Meta AI Support Assistant is designed to help you resolve account issues from start to finish. Get answers to all your questions, including notification settings and new features. You can also optionally take actions on increasing requests within Facebook and, in the future, on Instagram, such as:
- Report fraud, impersonated accounts, and questionable contentt
- Easily see why your content was removed, have options to appeal, and track what happens next.
- Manage your privacy settings
- Reset password
- Update profile settings
Getting support should be easy. The Meta AI Support Assistant is built into Facebook and Instagram, so help is always just a tap away. We typically respond to requests in less than 5 seconds, significantly reducing wait times compared to traditional Help Center searches or external websites for answers.
Meta AI Support Assistant is an important step in providing stronger support in your app. In fact, the majority of people who provided feedback reported a positive experience with Meta AI Support Assistant. Currently deployed in all supported languages. facebook and Instagram For support topics.
We’ve also started rolling out Support Assistant for people who need help logging into their Facebook or Instagram accounts. We’re starting with select cases in the U.S. and Canada, and will soon expand to more countries and other types of account access situations. We continue to invest in AI-powered tools to make support more accessible, reliable, and effective. We’ll also continue to evolve the Meta AI support assistant and continue to improve it over time as more people use it and technology advances. Learn more about Meta AI Support Assistant here.
Improved content enforcement with more advanced AI
Last year we shared positive results These are some of the changes we’ve made to reduce errors and focus on proactively cracking down on the most serious illegal content on our platform, including terrorism, child exploitation, drugs, fraud, and fraud. us too share We have been experimenting with more advanced AI systems for content enforcement to build on this progress. We believe this system will intercept more fraud and reduce over-enforcement errors, while capturing more violations more accurately and responding quickly to real-world events.
Initial testing of these systems promises to:
- It reduces the likelihood that fraudsters will trick people into divulging their login information, ultimately discovering and mitigating the 5,000 frauds per day that previously went undetected by existing review teams.
- Identify and prevent accounts that impersonate celebrities and other public figures. This allowed us to reduce user reports of the most commonly impersonated celebrities by over 80%.
- We catch twice as much violating adult sexual solicitation content than our review team. At the same time, the error rate is reduced by more than 60%.
- Prevent account takeover by recognizing when someone suddenly accesses your account from a new location, changes your password, or edits your profile. On their own, these changes seemed harmless to someone verifying the account, but the AI could see them as a threat.
- Detects fake sites masquerading as legitimate web addresses and masquerading as popular sporting goods stores by focusing on genuine logos and suspicious web addresses used at unusually low prices. After being tested more extensively, this AI delivered encouraging results, reducing views of ads with fraud and other serious violations by 7%, increasing user and brand protection.
These more advanced AI systems can do all this in the language spoken by 98% of people online. This is far more than the approximately 80 languages we previously covered. Importantly, you can develop your proficiency in any language based on your needs and adapt to understand cultural nuances, including niche subcultures, rapidly changing regional idioms, emoji meanings, and slang.
A smarter approach
Over the next few years, once we see that these more advanced AI systems consistently perform better than current content enforcement methods, we plan to deploy them across our apps and transform our approach. As we do this, we will reduce our reliance on third-party vendors for content enforcement and focus on strengthening our internal systems and employees. There will still be people reviewing content, but these systems will be able to do tasks better suited to technology, such as repeated reviews of graphic content and areas where adversaries constantly change their tactics, such as illegal drug sales and fraud.
Even as we use new technologies to expand our possibilities, people remain at the heart of our approach. AI can help us move faster and operate at scale, but it cannot replace human judgment. AI helps us apply judgment more consistently across billions of pieces of content on our platform. Experts design, train, supervise, and evaluate AI systems to measure performance and make the most complex and impactful decisions. for example, People will continue to play a key role in making the riskiest and most important decisions, such as deactivating accounts and reporting to law enforcement.
We rigorously test each of these AI systems, protect against bias, incorporate safeguards to ensure consistency and accuracy, and evaluate their performance. Community standards will not change as part of this transition. We’re also using new tools like the Meta AI Support Assistant to improve the way you report violating content and complain about mistakes. Ultimately, this approach also helps ensure that humans do what humans do best and technology does what technology does best. This means combining the scale and capabilities of advanced AI with human expertise and judgment, each reinforcing the other.
