Satellite imagery has carried unusual evidentiary weight precisely because it was expensive and hard to fake. Google briefly removed that property, then reversed within 24 hours. The rollback is the useful part of the story: it is a rare, documented case of a platform discovering the failure mode before the failure, and it does not solve the underlying verification problem.
Sometimes, but not reliably from inspection alone. Google states that images from its Google Earth feature carried SynthID watermarking and C2PA Content Credentials, checkable through the Gemini app or Lens in Search. Those are Google's claims about its own system and apply only to images from that source — a fabrication produced by a different model may carry no marker at all. The dependable approach is provenance rather than detection: trace the image to a named satellite operator and acquisition time, and cross-reference against independent imagery of the same location.
DrafterDaily explains what shipped, what broke, and what it means. One email each morning.
CVE-2026-0768 is an unauthenticated root RCE in Langflow. It was disclosed in January, the fix has shipped through seven releases, and attackers are hitting it in September — because low-code AI middleware became critical infrastructure without acquiring a patch owner.
Opposition to data centres is not a referendum on AI. It is a cost-allocation dispute — and PJM's capacity auction is the mechanism that turned an abstract argument into a number on 67 million households' bills.
OpenAI's first published benchmarks for its custom inference chip are real, from a public benchmark, and normalised on nameplate TDP rather than measured power. OpenAI disclosed that itself, in a sentence nobody is quoting.