Companies adopting artificial intelligence are running into a basic operational problem: they can measure AI usage in tokens, but they still struggle to predict what those tokens will ultimately cost. Because large language models are non-deterministic, the same prompt does not always produce the same response, and small changes to instructions or model choice can materially alter token consumption. That makes long-term budgeting difficult, especially as firms move from single-model deployments to systems built around multiple AI agents working together. ABMedia reports that this pressure is spreading across the technology sector. Goldman Sachs has forecast that global token consumption will rise 24-fold between 2026 and 2030, reaching 120 quadrillion tokens per month as enterprise adoption of AI agents expands. Some businesses only realize they have exceeded budget limits when monthly bills arrive; Uber, for example, was cited as having exhausted its annual token budget within a few months. To respond, companies are tightening controls. Measures now in use include restricting expensive external AI tools, training employees to write more precise prompts, choosing models based on fit and cost rather than novelty, and budgeting for security testing and safeguards that also consume tokens. Small firms using personal accounts to avoid higher enterprise pricing were also warned not to treat that workaround as a durable solution.

Suno says it plans to add watermarks to all audio generated by its models in the coming weeks, giving partner platforms a way to label or block AI-made tracks. The company did not say what technology it will use, did not provide a launch date, and did not disclose how download limits may change. The announcement came after a German court ruled on July 31 that Suno violated local music licensing rules, while Sony’s copyright case against the company in the U.S. also moved forward in July at the summary judgment stage. Suno said the watermark would embed an inaudible code in the waveform and would be durable, resistant to tampering, and not affect audio quality. The company said the goal is to help partners fight fraud and abuse. Still, the legal focus remains on how Suno obtained training data, not whether finished outputs carry labels. The report also points to a 2025 data breach that allegedly showed Suno scraped material from YouTube, Deezer, and Genius for model training, an incident that has since led to a class action lawsuit in Massachusetts.

Spotify reported its second-quarter 2026 results, with paid subscribers rising above 300 million for the first time after adding 7 million net subscribers in the quarter. Monthly active users reached 777 million, up 12% from a year earlier. Revenue came in at 4.78 billion euros, up 14% year over year, while net profit was 545 million euros. Gross margin hit a record 33.4%, and operating income of 655 million euros came in above the market expectation of 639.2 million euros. Still, the overall report and the company’s guidance for the third quarter fell short of market expectations in key areas. Spotify said it expects Q3 monthly active users to reach 788 million, below analysts’ estimate of 793 million. It also forecast operating income of 670 million euros, under the expected 677.8 million euros. Shares closed down 1.57% on Tuesday. The company said it is stepping up spending on marketing and AI, with second-quarter operating expenses rising 19% year over year to 941 million euros. Spotify also said it has partnered with independent music licensing agency Merlin to expand AI remixing and AI cover tools to more than 30,000 independent labels, with plans to launch the tools as paid add-on services.

A court in Munich ruled Friday that AI music company Suno infringed copyright and must disclose revenue derived from the infringement, according to Reuters. The regional court said Suno was not authorized to process songs represented by Gema, Germany’s statutory collecting society. The decision adds to a broader legal fight in which artists and publishers are challenging technology companies over the use of copyrighted works. Under the ruling, Suno must also pay damages, though the amount has not yet been determined. The judgment can still be appealed to a higher court.

A German court has ruled against AI music startup Suno in a copyright case brought by music rights group GEMA, finding that the company used music from GEMA’s catalog to train its models and reproduced protected works without a license. According to Variety, the Munich Regional Court said commercial use of GEMA’s repertoire by AI companies requires licensing, and that requirement covers both model training and music generation. The case focused on six songs, including “Daddy Cool,” “Rasputin,” “Forever Young,” and “Mambo No. 5.” GEMA argued that Suno trained its AI on copyrighted songs without permission. Suno rejected that view, saying its platform was designed to help users create new music rather than copy existing tracks, and said the ruling rests on a misunderstanding of how its technology works. The company said it is reviewing its legal options, including a possible appeal. The decision came about eight months after GEMA won a separate copyright case against OpenAI over ChatGPT’s reproduction of copyrighted song lyrics. It also lands as Suno faces wider scrutiny over its training data after leaked source code earlier this month pointed to more than 113,000 hours of YouTube Music, 62,000 hours from Pond5, and 12,000 hours from Deezer in its dataset.

Writers, illustrators and independent musicians are pushing a widening set of lawsuits against Google, Meta and Anthropic over how AI systems were trained, with disputes spanning copyright infringement, allegedly unauthorized use of platform data and the scope of user agreements. The legal picture remains mixed. Anthropic reached a $1.5 billion settlement with authors in a case approved on July 20, 2026 by San Francisco federal judge Araceli Martínez-Olguín, covering roughly 500,000 works at about $3,000 apiece, and agreed to destroy pirated ebook files. But the same litigation also left room for AI companies, after judge William Alsup found that training on lawfully purchased and scanned second-hand books under the so-called "Panama Project" qualified as fair use. Other cases are still moving through court. Artists have continued their long-running suit against Stability, Midjourney, DeviantArt and Runway AI; musician Sam Kogon is challenging Google over YouTube Content ID data allegedly used to train Lyria and ProducerAI; and claims against Meta over Llama were narrowed after a judge found insufficient evidence of market harm, while piracy-related allegations survived. New suits filed this year by music publishers, book publishers and authors show the fight is still growing.

AI music startup Suno has raised $400 million at a $5.4 billion valuation and reached a settlement and licensing deal with Warner Music. The company is testing new products that will let users create with Warner-owned music assets.

Sony Music’s copyright case against Suno is entering a summary judgment hearing in federal court, putting the legality of training AI on copyrighted recordings before a judge for the first time.