Elon Musk’s X Is Turning Group Notes Over to AI

Synthetic intelligence chatbots are identified for frequently providing doubtful data and hallucinated particulars, making them horrible prospects for the position of fact-checker. And but, Elon Musk’s X (née Twitter) plans to deploy AI brokers to assist fill within the gaps on the notoriously slow-reacting Group Notes, with the AI-generated notes showing as quickly as this month. What might probably go fallacious?

The brand new mannequin will enable builders to submit AI brokers to be reviewed by the corporate, based on a public announcement. The brokers will probably be examined behind the scenes, made to put in writing follow notes to see how they may carry out. If they provide helpful data, they’ll get the inexperienced mild to go reside on the platform and write notes. These notes will nonetheless need to get the approval of the human reviewers on Group Notes, and so they nonetheless should be discovered helpful by folks with quite a lot of viewpoints (how that metric is set is a bit opaque).

Builders submitting their very own brokers could be powered by an AI mannequin, based on Bloomberg, so customers gained’t be locked into Grok regardless of the direct ties to Musk (maybe as a result of Musk merely cannot stop Grok from being woke, regardless of how hard he tries). The expectation from the corporate is that the AI-generated notes will considerably enhance the variety of notes being revealed on the platform.

They kinda want the AI for that, as a result of human-generated notes have reportedly fallen off a cliff. An NBC News story revealed final month discovered that the variety of notes revealed on the platform was reduce in half from 120,000 in January to simply 60,000 in Might of 2025. There are fewer folks submitting notes, fewer folks ranking them, and fewer notes being displayed. Mainly, the engagement with the fact-checking service has collapsed.

There are possible quite a lot of elements for that. For one, the platform is sort of a shit present. A Bloomberg analysis discovered that it takes about 14 hours to get a word hooked up to a publish with false or deceptive data, principally after its main viral cycle passes. Disagreements amongst Group Notes contributors have additionally led to fact-checks failing to get revealed, and about one in 4 get pulled after being revealed as a result of dissent amongst raters. That determine will get even increased when associated to actively contentious points like Russia’s invasion of Ukraine, which noticed greater than 40% of revealed notes ultimately taken down.

After which there’s the man who owns the positioning who, regardless of actively selling Group Notes as a giant repair for misinformation, has spent an increasing number of time shitting on it. Earlier this 12 months, Musk claimed, with out offering proof, that Group Notes may very well be gamed by authorities actors and legacy media, instilling mistrust in your complete course of. You already know what isn’t going to make the system tougher to recreation? Unleashing bots on the issue.