OpenAI ships always-on GPT-6 Dots with 4,000+ app connections
Dots can act continuously in the background, have their own computers and can be granted access to yours, and the author urges companies to get permissions right before turning them on.
openAI shipped always-on GPT-6 Dots with 4,000+ app connections. AI engineering's biggest question now: who's watching them? dots are impressive. they also do things nobody has shipped at this scale before: → they act continuously, in the background, with minimal oversight → they have their own computers, and can get permission to use yours → they plug into thousands of apps: email, Slack, code, billing → they're designed to act before you ask openAI says Dots follow the same permissions and controls as ChatGPT, and you choose what needs your approval. still, several outlets noted the launch landed amid safety questions. the upside is huge. so is the need to get your permissions right on day one. i'm breaking down how to roll out Dots safely over the next few days, and the first one drops tomorrow. if your company is about to turn them on, show this to whoever owns security so they don't miss it. how much autonomy would you actually give an agent with access to your inbox?
This post quotes another post: x.com/teka1900/status/2105602481506156562