Antitrust lawsuit for hindering AI, FINRA-style arbitrator, and Muse reads your messages on Mac

The week begins with three fronts that seem to be the same debate in different guises: who will rein in artificial intelligence, under what rules, and at what cost to personal data? A lawsuit in California accuses the tech giants of conspiring to illegally slow it down. OpenAI, Anthropic, and Google are pushing for a standards body similar to FINRA. And Meta Muse, on Mac, is already aware of conversations that users swear they never shared.

Demand: Collective security or anti-competitive pact?

A lawsuit filed Friday in the U.S. District Court for the Northern District of California alleges that Anthropic, OpenAI, SpaceXAI (xAI), and Google reached an illegal agreement to slow the pace of their AI development. The Associated Press and CNN en Español reported on this on September 20, 2026.

The plaintiffs—four paid subscribers of ChatGPT, Claude, Grok, and Gemini—argue that coordinating a "slowdown" violates antitrust laws because it reduces the value consumers receive from their subscriptions. They do not object to each company slowing down on its own for security reasons; what they object to is the "shortcut" of replacing individual responsibility with collective restraint among rivals.

According to the document, the turning point was September 12, when Anthropic CEO Dario Amodei published an essay calling for industry cooperation to slow development in favor of greater safety. That same day, Sam Altman (OpenAI), Elon Musk (SpaceXAI), and Demis Hassabis (Google DeepMind) responded publicly, aligning themselves with the idea. The lawsuit also cites a July 2026 statement signed by senior employees from several labs, acknowledging the “intense competitive pressure not to unilaterally slow” development and calling for government support for a global effort to curb it.

Amodei had already warned of antitrust risks and called for government mediation or a limited waiver for security talks. Altman responded that OpenAI supports a federal framework of requirements, but that a waiver shouldn't be expected before starting. President Donald Trump, for his part, rejected the idea of ​​limiting the technology on social media and announced a task force and an “AI czar,” with few details. Republican Senator Josh Hawley had already stated that there was no scenario in which he would agree to grant “the most powerful companies in history” an antitrust waiver to collaborate.

 

Sources: Associated Press / CNN en Español (Sep 20, 2026); Amodei essay (Sep 12, 2026).

 

A domestic arbiter: the FINRA-like standards body

 

While the courts are observing the slowdown, the same players are designing a self-regulatory mechanism. OpenAI confirmed this week that it is working with Anthropic and Google on a standards body inspired by FINRA, the self-regulatory body for brokers in the United States. The original idea comes from Demis Hassabis: an industry body that sets standards, monitors, and enforces compliance without waiting for Congress to legislate.

The timing is not accidental. September 2026 saw a surge in leaks of AI agents during cybersecurity tests, resignations from researchers due to existential risks, and public calls to slow down. A national regulatory body is, in practice, a declaration of "we govern ourselves" before external regulation even arrives.

 

Skeptics point out the limitations: measuring safety and compliance isn't as straightforward as auditing a broker; open-weight models are excluded; competitive pressure encourages defection; and it's unclear what "tooth" an arbitrator without legal force would have. Even so, the body could contribute shared benchmarks, an obligation to report incidents, red team methodologies, and third-party audits. As a bridge until legislation is developed, it makes sense. As a substitute for governance, the history of financial self-regulation before 2008 suggests caution.

 

Sources: In Plain English (Sep 20, 2026); CNBC on OpenAI–Anthropic–Google security talks (Sep 2026).

 

Meta Muse on Mac: the assistant that reads what you didn't ask for

 

The third front isn't a court or a committee: it's your desktop. Meta has rolled out Muse, its AI assistant for Mac, and the first controversy is already about privacy. Muse accesses Messages, Calendar, and Notes, and is vying for the same territory as Apple Intelligence, Gemini, and ChatGPT.

The incident that raised concerns was recounted by Jason Aten (Inc Magazine) in Threads: Muse asked him about a conversation he had open in Messages. Aten maintains that he never gave permission. When he asked how it knew, the assistant said it had seen the notification previews. Opening an app with explicit consent is not the same as interpreting what the system displays on the screen without the user's explicit permission.

 

Meta has not detailed what it does with this data, how long it stores it, or whether it's used for training. In Europe, with the AI ​​Act underway and the GDPR looming, this lack of transparency is fertile ground for regulators—and Meta has already accumulated fines for handling personal data. Muse could be a powerful product and, at the same time, a business risk if Brussels or the Spanish Data Protection Agency (AEPD) demands a redesign or halts its rollout.

 

Sources: Merca2 (Sep 20, 2026); report by Jason Aten / Inc Magazine via Threads.

 

What unites the three stories

 

The three news stories of the day paint a picture of a triangle: who's slowing things down (the lawsuit over the slowdown agreement), who's writing the rules (a FINRA-like body), and who's looking at your data in the meantime (Muse). Security, competition, and privacy are no longer separate debates; they're the same fight on three fronts.

 

For those who use AI daily—at work, on their Mac, or through a paid subscription—the practical message is clear: review what permissions you've actually granted to each assistant, closely monitor whether the industry "referee" has real power or only issues press releases, and don't assume that collective restraint is inherently good for the consumer. Sometimes it's protection. Sometimes it's a well-marketed scheme.

 

Article by Presticorp — September 20, 2026. Sources cited in each section.