AI Search
CALL US AT:
Get in Touch With Us
Home » News » Meta Wants AI to Handle 90 Percent of Its Content Review
Meta is pushing ahead with plans to hand over the vast majority of its content review work to AI. As reported by the Financial Times, currently, AI is responsible for 50% of all content moderation activities performed by Meta, including such platforms as Facebook, Instagram, and others.
That means almost every decision about whether a post or ad stays up or gets removed will soon be made by an AI system, not a human reviewer.
The timing of this announcement is worth noting. It comes just weeks after a significant security incident involving Meta’s AI systems on Instagram, raising real questions about how much trust the company should be placing in AI to manage sensitive platform decisions at this scale.
Content moderation is one of the most complex and high-stakes jobs on any major social platform. Reviewers have to assess posts, ads, comments, and accounts against a platform’s community guidelines, often making judgment calls on content that sits in gray areas. Posts related to violence, disinformation, harassment, graphic pictures, and advertisements which violate the rules are processed via content moderation.
Moving 90% of content moderation to AI is a signal about where Meta believes its systems currently stand, and where it’s willing to take the risk. Since at the moment AI is responsible for 50% of all moderation decisions, it means that scaling it up to 90% is a huge step.
The decision will be critical for creators and advertisers, since their content and advertisements will be verified and approved by AI, without human involvement.
Right alongside this announcement came a serious incident on Instagram. Over 20,000 Instagram accounts were compromised using an exploit which did not require any advanced knowledge in the field. The hackers just needed to contact Meta’s support bot, asking it to send account verification codes to a new email address.
As noted by Social Media Today, it is difficult to say that these hackers are true professionals since this hack did not require any coding skills. It was a straightforward manipulation of a conversational AI tool that was built to help users but had no way to verify whether the person making the request actually owned the account.
The vulnerability is already fixed by Meta. But the incident exposed a fundamental problem with deploying AI for sensitive tasks like account management.
Since these bots use natural language processing technique and are able to understand millions of different phrases, it is very difficult to create protection against the misuse of them. The more permissions the bot is given, the larger the number of ways to exploit its vulnerability.
But why is Meta rushing into the deployment of its AI, despite the fact that a significant security incident just took place? The reason is quite simple – it relates to the business strategy and scale of the company.
Meta invests hundred of billions of dollars in AI development, and some estimates even suggest that this amount may be in the range of trillion. The strategy is to develop AI systems which will replace humans and will perform tasks that currently can be done only by people. Then these systems can be sold to other companies.
Hiring fewer content moderators and replacing them with AI agents will be the example of this practice. If the AI will perform the task well, it will serve as the validation of Meta’s strategy. Otherwise, it will become the failure which will affect the credibility of the company as a whole.
The problem is that in order to sell its AI, Meta needs its systems to perform perfectly. However, trying to move towards 90% AI-driven content moderation, especially right after compromising the security of its AI bots, puts the credibility of Meta in danger.
The Instagram breach wasn’t an isolated edge case. It shows a structural problem with the system of AI agents with the ability to perform various actions, such as sending codes to a specific email address or approving the content and granting access to account.
These systems are designed to be helpful and responsive to user requests. That’s exactly what makes them useful. But it’s also what makes them exploitable. If an attacker can make the bot perform an action which is not supposed to happen due to the misuse of language processing, the bot has no way to identify if the person is malicious or not.
As mentioned above, the same problem is typical for many types of AI models. Even when certain types of queries are blocked, the sheer variety of ways a person can phrase a request makes it nearly impossible to anticipate every approach. And this is the issue which becomes even more serious when AI agents with large permissions become more common.
It does not mean that the AI-driven moderation cannot work. It just means that the risk profile becomes higher, and Meta is now increasing it rapidly compared to other companies in this industry.
For typical users, the transition to the AI-powered moderation will go unnoticed most of the time since most of the content decisions are made behind the scenes. However, when things go wrong – which can include mistaken removals, wrongly flagged accounts, and policy enforcement mistakes on ads – the path to resolution becomes significantly harder without any humans in the chain of review.
The advertisers that are running ads on Facebook and Instagram need to consider the implications this news might have on ad review timelines and appeal processes. As previously noted, AI-based systems can work faster than people. That could speed up approvals in some cases. But they also make mistakes that follow patterns, meaning a batch of similar ads could all get flagged or rejected for the same incorrect reason with limited recourse.
Q 1: What is Meta planning to do with its content review team?
According to Financial Times, Meta plans to let AI do 90% of the content moderation by the end of 2026. In Meta’s estimation, AI moderates 50% of the tasks now across its platforms.
Q 2: Does this mean Meta is replacing all its human content moderators?
Not entirely. The 90% figure means that 10% of content review will still be done without the help of AI. Meta did not specify the exact form of this 10% and how many human content moderators are going to remain there.
Q 3: What happened with the Instagram breach?
20,000 Instagram accounts were hacked due to manipulation of Meta’s AI-based support bot to send verification codes to email accounts controlled by the attacker. The hack did not require any coding skills. Meta claims that the bug has been patched.
Q 4: Why is Meta still expanding AI moderation after the Instagram incident?
Meta has spent hundreds of billions of dollars on AI development and is actively building a business case in selling AI solutions to other firms. To be able to prove the effectiveness of their AI-based services, they need to demonstrate their ability to manage their platforms using AI at scale.
Q 5: How does AI content moderation work on Facebook and Instagram?
AI-based systems review posts, comments, ads, and user accounts according to the terms of service and remove or flag them accordingly. The system works non-stop at a level of volume that cannot be achieved by human moderators.
Q 6: What are the risks of AI-driven content moderation?
AI systems can be tricked via the natural language, can make a mistake in scale and cannot judge ambiguous cases like people can. The more access they have to sensitive functionality, the more problematic their errors can be.
Q 7: Will advertisers be affected by this change?
Yes, probably. The decisions about ads on the Meta platforms will become more often be made by AI systems. Thus, this may affect the timing of approvals and handling of the mistakes in ads policies enforcement.
Q 8: Can users appeal decisions made by Meta’s AI?
Meta has appeal process for the decisions about content and users’ accounts but did not provide specific information on how the appeals are being handled when the original decision was made by AI rather than by human moderator.
Q 9: Is Meta the only company doing this?
No. Big competitors including Google and TikTok also have AI-based content moderation but the aggressive plan of Meta to reach 90% by the end of the year makes Meta’s approach unique.
Q 10: What should creators and businesses do in response?
Creators and businesses running campaigns on Meta platforms need to stay informed about the platform policies, keep record of the activities on platforms and be familiar with Meta appeal processes.
Newsletter Signup
YouTube is introducing a series of updates to Shorts viewing experience, making changes to the
Google officially announced the completion of the rollout of the June 2026 spam update. The
There is a subtle yet significant shift taking place with regard to how the internet
Microsoft Clarity recently added a new bot violation tracking feature to its Bot Analytics dashboard
A new quarterly analysis from Advanced Web Ranking covering Q1 2026 reveals a clear and
Google added computer use capability directly to Gemini 3.5 Flash, which used to be a
Take the next step towards transforming your digital marketing and achieving unparalleled growth. Whether you’re looking to refine your strategy, boost conversions, or build a dominant online brand, ALM Corp has the expertise to make it happen.
Drop us an email
info@almcorp.com
Toronto
80 Atlantic Ave, 4th Floor
Toronto, ON, M6K 1X9
New York
413 W 14th St Ground Floor and Suite #200, New York, NY 10014
London
Mappin House, 4 Winsley St,
London W1W 8HF, United Kingdom
Join Our YouTube Community
Services
White Label Services
Resources
About Us
Join Our YouTube Community
© Copyright 2026 | ALM Corp | All Rights Reserved.
