September 8, 2026
TechCrunch published a glossary this week walking through terms like "opaque recurrence" and other AI jargon that's crept into boardrooms without anyone pausing to define it. That sounds trivial until you sit in a vendor call and realize the salesperson, your CTO, and your CFO are all nodding along to the word "agent" while picturing three different things. This isn't just an industry quirk -- it's a real business risk. If your team can't agree on what "hallucination," "fine-tuning," or "context window" actually mean, you can't evaluate a vendor's claims, you can't write a sane contract, and you definitely can't hold anyone accountable when a tool underperforms. We've argued before that this vocabulary gap is a genuine barrier to adoption, not just an annoyance, and it's why we maintain our own glossary for exactly this reason -- shared language is the first step to a shared understanding of risk.
TechCrunch also reported that authors are pushing back against publishers and literary agents who they say are claiming an outsized share of the Anthropic copyright settlement. This is worth watching closely, because it previews a fight that's coming to every industry AI has trained on. The settlement exists because Anthropic used copyrighted books to train its models -- a fact the company has effectively conceded by paying out. But writing the check is the easy part. Deciding who gets paid, and how much, is where the actual power struggle happens, and right now it looks like the parties with the most leverage (publishers, agents) are positioned to out-negotiate the individuals whose work was actually used. If you're a business relying on AI tools trained on scraped or licensed content -- marketing copy, code, images -- this is a preview of the liability and payout disputes that could eventually touch your own vendors. It's a good reminder to actually read the data and IP terms in your AI contracts rather than assume someone else settled it for you.
TechCrunch reported that Travis Kalanick's company Atoms is reportedly exploring a move into robotaxis, with Kalanick describing it as a chance to finish what he started at Uber. Whatever you think of Kalanick's Uber history, the instinct here is telling: the founder who built a ride-hailing empire on human drivers now wants back in, this time on the automation side. That's a signal about where the money in transportation is actually flowing, and it's consistent with the flood of AI infrastructure and robotics investment we've tracked across the industry over the past few weeks. For most businesses this isn't directly relevant -- you're not launching a robotaxi fleet -- but it's a useful data point on how aggressively founders with capital and name recognition are betting on AI-driven automation displacing labor-heavy business models. If it can happen to ride-hailing, it's worth asking which parts of your own operation look like Uber circa 2010.
All three stories are really about control: control over language, control over who gets paid when AI uses your work, and control over which industries get automated next. None of these are settled questions, and the companies moving fastest right now are the ones deciding the answers for everyone else. If you're building on AI tools, it's worth asking early who defines the terms of your contract, your data, and your workflow -- because right now, it's rarely you. We've written before about making sure you own your own app and its decision rights rather than ceding that control to a vendor by default, and this week's headlines are a reminder that the same logic applies far beyond software.
If your business relies on an AI vendor's data, training set, or automation roadmap, do you actually know who controls the terms you're operating under -- and what happens if that vendor's priorities shift?
Sources