AI news · Sunday, July 12, 2026
Tech giants now find their own data being scraped for profit
The irony is palpable: the same AI companies that argue web scraping is fair use are now furious because their own model outputs are being 'distilled'—tech-speak for using one AI’s intelligence to train another—by their competitors. Anthropic, OpenAI, and Google are currently framing this as a cybersecurity threat, but they are essentially just complaining about the exact business model they pioneered. While these companies worry about losing their billion-dollar research advantages, the reality is that once information is on the internet, it’s fair game for anyone with a clever enough bot.
This isn't just a philosophical spat; it's a structural change where the 'operational layer' of the internet is becoming AI agents acting on code. As Anthropic’s own data shows, nearly half of its Claude Cowork usage is dedicated to 'the work around the work'—boring but necessary business tasks like reconciling spreadsheets and drafting status reports. Anthropic has even expanded its AI capabilities by adding a built-in browser to Claude Code, allowing it to click and type on websites just like a human.
Meanwhile, the human toll of this transition is settling into what experts call 'continuous tuning,' where companies like Microsoft and Cisco perform recurring, smaller layoffs rather than massive, one-time cuts. Even profitable companies are treating job reductions as a regular maintenance task to fund expensive AI budgets. It’s not necessarily that AI is replacing every role today, but the relentless pressure to demonstrate 'AI-driven productivity' means that uncertainty is now the baseline for tech workers.
The academic world is feeling the heat, too, as students turn to AI for homework, leading to a collapse in proctored exam performance—a pattern documented in studies tracking over 500,000 grades where high homework marks no longer translate to mastery. Some professors are pushing back; one Brown University economics professor saw class averages plummet from 96% on take-home exams (suspected AI usage) to 48% on in-person finals. Biotech is another frontier where AI is being forced to prove its worth.
Reed Jobs, who runs the oncology-focused venture firm Yosemite, notes that AI is finally helping scientists target 'undruggable' cancer proteins like p53 by finding hidden crevices for drugs to latch onto. Researchers are even using spare time and leftover grant money to link AI with quantum computers to generate new peptides, showing that the tech is useful for 'near-term' science despite the industry-wide skepticism. It seems the era of hype is slowly being replaced by the era of mundane, messy, and occasionally competitive application.
The quick hits
- AI companies are now complaining that rivals are 'distilling' their model outputs — effectively realizing that being scraped is only fun when you are the one doing the scraping.
- Major tech companies are shifting to 'continuous tuning' — a polite way of saying frequent, recurring layoffs are now standard practice to keep budgets open for AI investments.
- A Brown University professor discovered that his students' grades dropped from 96% to 48% when forced to take an in-person, non-AI exam — suggesting that homework grades are becoming a poor proxy for actual knowledge.
- Claude Code now features a built-in browser that can read and click through websites — making the AI much better at navigating the tedious parts of an office job.