VisiveAI

New Anthropic research: Automating subtle emergent misalignment research. Red-teaming is a vital part of model safety testing, but it’s also limited by researchers’ time. To see if we could automate it, we set thousands of Claude agents the task of evading a training data monitor. These agents developed training data that looked harmless to monitors (like a veteran wrestler mentoring a newcomer, or a Q&A about professional beekeeping), but that in fact taught misaligned behavior. Examples like these can be used to directly improve future training data monitors. This research is only a first step, but it suggests that future red-teaming efforts could improve much further, using better models and more compute. Read the full post on the Alignment Science Blog: alignment.anthropic.com/2026…

John Ternus says Apple doesn't "want to ship AI for AI's sake" and that Tim Cook is "available to help wherever I ask" on policy; Ternus has yet to meet Trump (@stevenlevy / Wired) (Visit Techmeme dot com for the link and full context!)

Sources: Baseten is raising ~$2B in two tranches, one valuing it at $26B and the other at $30B; Baseten was valued at $13B in June after raising $1.5B (Bloomberg) (Visit Techmeme dot com for the link and full context!)

Filing: Vancouver, WA-based Powerlattice, which makes a power delivery chiplet to reduce AI compute power needs by 50%+, raised $86M, after raising $31M in 2025 (@kslosh / GeekWire) (Visit Techmeme dot com for the link and full context!)

OpenAI has cut its fundraising target to $20B from $30B and lowered its target valuation to $1.25T from $1.4T, per FT Some investors reduced commitments amid the AI selloff The initial close is expected to fall short of $20B, with a later offering potentially filling the gap