AI news · Friday, August 14, 2026

Anthropic AI watermarking sparks user exodus as subscriptions hit the chopping block

Anthropic is now embedding invisible watermarks in Claude's text outputs to satisfy new EU AI regulations. The company insists these markers are undetectable and don't change the writing quality, but the developer community isn't having it. Freelance developers and consultants are canceling their monthly subscriptions, citing privacy fears and concerns that their professional work—coding and academic research—could be traced back to a machine. Industry analyst Ben Thompson added to the noise, calling the concept of watermarking proofreading or ideation 'clearly absurd,' noting that it essentially forces a tool to claim credit for human thought. While Anthropic plans to release an API for verifying these markers, the backlash highlights a growing tension between bureaucratic compliance and the workflows of people who actually build things with these models.

Financial markets are showing that, at least for now, cheaper AI doesn't mean less profit. OpenAI recently slashed prices for its top-tier models by up to 80%, and rather than a revenue dip, the firm saw a massive surge in usage. Developers are following the 'Jevons Paradox' logic: when AI tokens become cheaper, companies stop hoarding them and start automating everything in sight. This trend has kept the investment frenzy alive, with Anthropic’s secondary-market valuation hitting a staggering $1.5 trillion. Even so, legendary investor Michael Burry is sounding the alarm, comparing the current AI infrastructure spending to the dot-com bubble and warning that companies are hiding massive risks through off-balance-sheet structures.

Infrastructure providers are finding that being at the center of the AI gold rush carries its own baggage. Hyperscalers like Amazon and Google are doubling down on natural gas to feed their data centers, but a new forecast suggests those energy bets might backfire. Gas prices in the US could triple due to this surge in AI-driven demand, turning what was supposed to be a stable power source into a volatile financial liability. Meanwhile, in the hardware space, French startup Kog is trying to prove that you don't need exotic new chips to win; they are betting that software optimization can squeeze enough performance out of the standard GPUs enterprises already own to make inference 30x faster. Even the chip giants are pivoting, with Nvidia incorporating Groq’s technology into its own ecosystem to neutralize the threat of faster AI processing. It seems the real race isn't just for smarter models, but for the most efficient way to keep the lights on and the chips running without breaking the bank.

The quick hits

Anthropic users are canceling their subscriptions because they don't want AI watermarks in their professional output — a sign that 'transparency' feels a lot like 'contamination' to power users.
OpenAI’s aggressive price cuts triggered a 14-fold jump in usage, proving that AI is following the classic tech adoption curve where lower costs lead to explosive, revenue-positive growth.
Hyperscalers are rushing to build massive natural gas plants to power their AI ambitions, but experts warn that surging demand could cause energy prices to triple, creating a massive financial risk.
Secondary markets have valued Anthropic at $1.5 trillion, yet shares remain nearly impossible to buy because investors are betting the company's upcoming IPO will be a historic payday.

Anthropic AI watermarking sparks user exodus as subscriptions hit the chopping block

The quick hits

Sources