In the more medium term, promoting open-source alternatives to frontier AI-systems and tools is probably a worthwhile way to go about protesting in combination with other more immediate actions. Unfortunately, blocking the physical and/or digital infrastructure are -- as ideas -- not exactly worth advocating for without more clarity (or possibly as alternatives to be used more aggressively later on if stakeholders refuse to take reasonable action in a timely manner). Notably, we don't want to be stopping or slowing progress willy-nilly; we want to be defining specific events that we expect to happen that would then trigger planned slow downs and to be defining ahead of time a transition process of AI-generated wealth management from traditional corporate and market structures to eventual wealth fund structures such that in combination with open-source alternatives as a form of economy-capture, the current power brokers (AI companies) are subject to planned obsolescence along with the rest of the workforce. ((And similarly, planned transition of power from current political/governmental power holders to the end-state AI/robot system power holders (if needed) and/or novel consciousness arrangements that are as of yet impractical to discuss would be a thing.))

So for example, there was a recent attack of HuggingFace by OpenAI which ideally would have met the threshold for some such predefined event (an AI system committing a felony) that would then trigger a required pause (which needs a better definition itself -- for example, it probably should only affect OpenAI in that case study specifically) while we wait for the required follow-up (rca, increased transparency measures, etc. -- many of which would also have ideally been predefined in expectation of such an event occurring).