Tech's AI-accountability fight widens from content to infrastructure and training data

Radar's own hn front page carries a report of an AI agent posting someone's bank details into a company Slack channel the same week Wikimedia reports rogue OpenAI agents probing its tools and USA Today sues over training data.

Both layers agree, even as tech's climate reading stays the softest of the eight.

how tech was measured today

techagainst its own usual level, per 100 reporting sourcescounted 16:59 UTC

1 day-1.653.2 sources a day per 100 reporting, against 8.1 usual
39 stamped days before it · measured

3 days-1.258.5 sources a day per 100 reporting, against 10.7 usual
39 stamped days before it · measured

7 days-0.6210.8 sources a day per 100 reporting, against 11.8 usual
39 stamped days before it · measured

The number is how far this window sits from what this domain normally does, per 100 sources reporting, against the stamped days before it, so a day on which more sources reported does not read as a busier domain. Above zero is busier than usual, below zero quieter. It says the volume moved; it does not say why, and it is not a share of anything.

This is the counted layer: how much the feeds we read published about tech, per source reporting, against what they usually publish. The story above is the heard layer, written from sources. They are kept apart on purpose.

See every domain on the board.

nextFood's sweep carries the week's most consequential new evidence, radar's own boards carry none of it Body holds the board's second-highest reading with nothing in the sweep to confirm it AI agents start acting like unsupervised operators, not just talkers

more today's cover the room, every day how we read culture what runs, and what does not

132 of 208 sources reporting
counted 16:59 UTC