It can be “easy” depending on your definition of easy. For the average person, no, not easy. For a tech person, maybe, depending on their interests and proficiencies. Easiest is likely using a firewall on the router blocking all outbound and inbound WAN connections (for the MAC address associated with the device, assuming it doesn’t try to rotate MAC addresses or hop on any open WiFi networks around you). Some devices will straight refuse to work if you do this, however. So you have to do research and be careful about which products you buy.

If you want a vacuum, see the valetudo link I posted in another comment. It has some options for this that you can load local only custom firmware (and integrate it with home assistant).

You can have local control over such things, and IOT devices should be locked down to only ever access via LAN anyways.

https://valetudo.cloud/

Don’t buy IOT devices you can’t control. If it’s not usable via valetudo or is offline capable don’t get it. Tell your friends and family the same. Own your technology, limit your attack surfaces.

I think nekokoneko above has it right, trying to misattribute blame to the user. Basically, he wants to blame the user for using the service, instead of fixing the service so it doesn’t produce heinously horrible things on request. It shouldn’t work, but who knows with the amount of money he has and the lack of money on the other side. The goal is to establish a small precedent, though if someone sues xAI for making a product that makes this kind of refuse, I doubt such a precedent would hold much water.

No, because when they do the training and actual updates to the model weights, they are very likely heavily favoring their internal renditions of preferred response to certain queries. So even if they included such things from online scraping, they’d essentially wipe the sentiment with internal overrides (most likely, LLM models are very big, you could absolutely get it to say almost anything verbatim it’s seen in the training data if you tried).