AI abundance is producing three different queue rules
LinkedIn separates user reports, Snap changes recommendation eligibility, and Apple limits security submissions. The three approaches show how low-quality AI output strains different review queues across platforms.
Artificial Intelligence··Morning
LinkedIn gave the problem its own reporting category
LinkedIn added a reporting option for low-quality AI content that sits apart from its standard content-reporting flow. A user can therefore flag the production quality directly without first claiming that the post violates another rule. The change does not make an automatic judgment about the content; it takes an initial classification from the user and sends it into the review process as a distinct signal. THE DECODER's report does not say how many reports have used the category, what share resulted in removal or reduced reach, or what feedback the person reporting receives. It also provides context about the scale of AI production on short-video platforms: a Kapwing study classifies 21 percent of YouTube Shorts as AI-generated, although that figure does not measure LinkedIn's feed. The separate button is consequently an observable interface change showing that the platform has named the problem. Whether it reduces the amount of low-quality material users see cannot yet be evaluated through a published denominator and outcome series. The mechanism here collects human signals in a separate review queue rather than blocking content before it appears.[1]
Snap separates visibility by production method
The same report says Snap has chosen a different instrument for Spotlight. This month the platform is removing wholly AI-generated videos from recommendations, while material made or enhanced with Snapchat's own AI editing tools remains eligible and carries a transparency label. The rule operates at the distribution stage without waiting for a user report, and it separates content hosting from recommendation visibility. Snap says the number of people contributing to Spotlight has grown by more than 120 percent, but it does not disclose how much of that growth comes from AI content. The report also gives no count of videos dropped from recommendations or their previous share of total viewing. The direction of the change is therefore clear, but its magnitude is not. LinkedIn's separate report option adds a user signal to a review queue; Snap's rule keeps a defined production method out of a recommendation queue. Both intervene in the visibility of low-quality or wholly synthetic material, but at different moments and with different inputs: one depends on user classification after publication, while the other depends on the production method before recommendation.[1]
Apple placed a direct quota on the security queue
Apple's security bounty problem concerns technical review capacity rather than public visibility. According to a report citing the Financial Times, low-quality AI submissions containing invented vulnerabilities clogged the queue, after which Apple limited researchers' reports and imposed a 30-day wait. The Italian startup Bynario says it found a macOS vulnerability with ChatGPT that could allow full control of a machine but could not submit it after hitting the cap; Apple later contacted the company. Meanwhile, Apple has used Anthropic and OpenAI models to hunt vulnerabilities, and its latest updates contained five times the usual number of fixes. The report gives neither the cap's start date nor counts of quality rejections or criteria for a higher quota. All three platform moves assume AI output is growing faster than review capacity, but intervene differently: LinkedIn classifies signals, Snap changes distribution eligibility, and Apple limits intake. Comparing outcomes would require action rates for reports, the share dropped from recommendations, and delays to genuine vulnerability submissions. The reports describe the rules and one side effect, not those outcome series.[2], [1]
Related columns
For more information on this topic, you can read the related columns.
