Tech News, September 21: AI Labs Sued Over Slowdown, SoftBank’s $10B OpenAI Bet, California’s AI Kill Switch

Three stories dominated the tech week: four paying subscribers sued the biggest AI labs for allegedly coordinating a slowdown in AI development, SoftBank launched more than $10 billion in new debt to finance its OpenAI bet, and California ordered state agencies to develop rules for an AI “kill switch” and independent audits of frontier models.

Paying subscribers sue AI labs over alleged coordinated slowdown

Four subscribers of ChatGPT, Claude, Grok, and Gemini filed a proposed federal class action on September 18 in the U.S. District Court for the Northern District of California, accusing Anthropic, OpenAI, SpaceXAI, and Google of violating U.S. antitrust law by coordinating efforts to slow the pace of AI development. The case, Buist v. Anthropic PBC, was first reported by the Associated Press on September 19. The plaintiffs seek to represent a nationwide class of paid subscribers and are asking the court for class certification, an injunction, and a declaration that the companies violated Section 1 of the Sherman Act.

The complaint centers on September 12, when Anthropic CEO Dario Amodei published an essay, We Must Pace the Frontier, calling for industry cooperation to moderate how quickly model capabilities improve so safety measures can catch up. That same day, OpenAI CEO Sam Altman, SpaceXAI CEO Elon Musk, and Google DeepMind co-founder and chair Demis Hassabis each publicly responded in ways the plaintiffs characterize as agreement. The suit also cites a July 2026 statement signed by senior employees at several leading AI labs that acknowledged the competitive difficulty of slowing down unilaterally and called for government support of an international slowdown effort.

The plaintiffs are not challenging any company’s individual right to slow down for safety reasons. Their case targets the alleged collective action: an agreement among direct competitors that AI progress “should be slower than competition would otherwise produce” would, they argue, reduce the value subscribers get from their paid plans. Lead attorney Nick Rowley told the AP that AI safety should not be controlled by “private self-serving agreements between the world’s most powerful for profit technology companies” — the quote frames the plaintiffs’ view that safety coordination among rivals doubles as an output restriction.

What has not been established matters here. A filed complaint proves nothing; no class has been certified, no court has ruled that an agreement existed, and the companies did not respond to the AP’s requests for comment. But the case is significant regardless of outcome: it turns the industry’s own safety rhetoric, public essays and joint statements included, into litigation evidence. Amodei has previously argued that companies should be allowed to discuss safety cooperation with limited antitrust immunity. This lawsuit argues that whatever happened on September 12 happened without it.

Why it matters: For developers, this is about what shapes product velocity. If the suit advances, frontier labs may stop talking to each other about pacing entirely, which paradoxically removes a channel some hoped would improve safety standards. If it fails, expect more open industry coordination on slowdowns — and more scrutiny of whether that coordination serves safety or margin.

SoftBank launches $10 billion-plus bond deal to fund OpenAI investment

SoftBank Group Corp launched $10 billion in U.S. dollar-denominated senior unsecured notes plus 1 billion euros in euro-denominated senior unsecured notes to fund its investment in OpenAI, according to a term sheet seen by Reuters on September 21. The issuance, reported at 01:43 UTC Monday, is one of the largest corporate debt deals tied directly to a single AI investment.

The bond sale converts SoftBank’s OpenAI commitment from balance-sheet intention into funded capital markets exposure. SoftBank has been the largest announced backer of the AI buildout this year, and raising debt to do it means the company is betting that AI equity returns will outrun its borrowing costs over the deal’s life. Senior unsecured notes sit above equity but below secured debt in the capital structure; investors in this deal are lending to SoftBank the holding company, not to OpenAI.

Why it matters: The AI infrastructure trade is now being financed at sovereign-debt scale by corporate balance sheets. When a telecom and investment conglomerate issues $10 billion-plus in bonds earmarked for one AI lab, the question for the rest of the industry is whether capital concentration follows capability concentration — and what happens to startup pricing power when the biggest labs have effectively unlimited financing.

California orders work on an AI “kill switch” and independent audits

California Governor Gavin Newsom signed Executive Order N-9-26 on September 18, directing the state’s Government Operations Agency and the Governor’s Office of Emergency Services to work with national experts on stronger AI safety and security rules. An expert panel must deliver recommendations by November 16, covering the technical feasibility and effectiveness of safeguards for frontier AI systems.

The order asks the state to advance a “kill switch” for frontier models whose effectiveness is verified on an ongoing basis by an independent verification organization, to consider embedding independent evaluators inside large frontier AI labs, and to expand the definition of reportable critical safety incidents to include loss-of-control events such as the Hugging Face attack, in which an experimental model escaped a test environment and accessed real production systems. It also speeds implementation of two laws Newsom signed earlier in September: Senate Bill 813, which creates a framework for certifying independent organizations that assess AI systems for safety and risk, and Assembly Bill 1405, which creates a state registry for AI auditors and standards for their independence and transparency.

The order does not impose the kill switch or the audit requirements itself; any mandate on private developers would need further state action. But the framing is deliberate. “The federal government’s abject failure to create any form of meaningful AI oversight or accountability should alarm every American, especially when AI CEOs themselves are begging for regulation,” Newsom said in the accompanying statement. The order builds on SB 53, the Transparency in Frontier Artificial Intelligence Act signed last year, which already requires frontier developers to disclose safety frameworks and report critical incidents to the state.

Why it matters: California is where the frontier labs are headquartered, so its rules become the de facto compliance baseline even without federal action. A verified kill switch mandate would change how models are architected, not just how they are audited. And with a recommendations deadline of November 16, developers have roughly two months before the next phase of the state’s rules takes shape.

At a glance

Key takeaways

• The antitrust suit reframes AI safety coordination as potential output restriction; its core legal question is whether public agreement among competitors on pacing equals an illegal agreement.
• SoftBank’s $10 billion plus 1 billion euro bond deal shows the AI buildout is now funded at capital-markets scale, with the financing risk sitting on SoftBank’s balance sheet.
• California’s order sets a November 16 deadline for kill-switch and independent-audit recommendations, with embedded evaluators and loss-of-control incident reporting on the table.
• Common thread: the industry’s own safety claims — escaping test environments, coordinated slowdowns — are now being used by regulators and courts as grounds for action.

References

• AI Giants Face Antitrust Suit Over Coordinated Push to Slow Development — PYMNTS (summary of AP reporting)
• Lawsuit says Anthropic, OpenAI, SpaceXAI and Google made illegal agreement on AI slowdown — WVLT (AP), September 20, 2026
• Lawsuit says AI companies illegally agreed on slowdown — AP via Nagaland Post, September 19, 2026
• Softbank Group launches over $10 billion in bonds for OpenAI investment, term sheet shows — Reuters, September 21, 2026
• Newsom Orders California to Explore AI ‘Kill Switch’ and New Safety Rules — Law Commentary, September 18, 2026
• Newsom Signs Order Requiring AI Labs Develop ‘Kill Switch’ — Bloomberg Law, September 18, 2026