News-May-23-2026

May 23, 2026 — House oversight of companies using Chinese-developed AI models drew new attention as lawmakers examined Airbnb and Anysphere's deployment choices, the NTSB restricted public docket access after AI-enabled cockpit-voice reconstructions, and a federal court's dismissal of Elon Musk's claims against OpenAI marked a major procedural endpoint in the AI industry's highest-profile governance dispute. A fourth update covered reports that the White House delayed an AI executive order centered on voluntary model-sharing and safety-testing after objections from technology executives.

Contents

1. House panels probe Airbnb and Anysphere over Chinese AI models
2. NTSB restricts docket access after AI voice reconstructions
3. Advisory jury rejects Elon Musk's OpenAI claims as untimely
4. White House delays AI executive order after safety-testing dispute

---

House panels probe Airbnb and Anysphere over Chinese AI models

Republican-led House committees investigated Airbnb and Anysphere, the maker of the Cursor AI coding platform, over the companies' use of artificial-intelligence models developed by Chinese companies. The House Homeland Security Committee and the House Select Committee on the Chinese Communist Party sent letters to the companies' chief executives requesting details about their use of Chinese-built AI systems, the reasons for those choices, and communications with the model providers. The letters, signed by Homeland Security Chairman Andrew Garbarino and China Select Committee Chairman John Moolenaar, also requested in-person briefings from employees involved in the decisions.

The inquiry targeted Anysphere's Composer 2 model after the company disclosed that the model was built on Kimi, a model developed by Beijing-based Moonshot AI. Lawmakers also questioned Airbnb's use of Alibaba's Qwen model for customer-service tools after Airbnb CEO Brian Chesky described the model as "fast and cheap." The oversight letters framed those deployments as part of a broader national-security and cybersecurity debate over whether U.S. firms' use of China-linked AI models can expose data, supply-chain, or model-behavior risks.

The investigation followed broader congressional and White House concern about alleged distillation campaigns by Chinese AI companies, including Anthropic's February allegation that Chinese-based AI companies used about 16 million exchanges from roughly 24,000 fraudulent accounts against Claude and OpenAI's letter to House China Select Committee members describing evidence indicative of DeepSeek attempts to distill frontier U.S. models. The development is significant for AI law because it links enterprise model procurement, data-security concerns, and congressional oversight of cross-border AI supply chains.

---

NTSB restricts docket access after AI voice reconstructions

The National Transportation Safety Board made its public accident-investigation docket system temporarily unavailable on May 21, 2026, while it reviewed materials that could allow reconstruction of cockpit voice recorder audio from sound-spectrum imagery. Ars Technica reported that internet users had used software and AI tools to recreate voices from the final seconds of UPS Flight 2976, a November 2025 crash that killed three pilots and people on the ground.

The NTSB said federal law prohibits public release of cockpit voice recorder audio and that advances in image recognition and computational methods had enabled individuals to reconstruct approximations of cockpit audio from imagery released in accident dockets. The episode is significant for AI law because it shows how generative and computational tools can convert lawfully released technical records into restricted biometric or voice-like material, prompting a federal agency to change public-access practices.

---

Advisory jury rejects Elon Musk's OpenAI claims as untimely

A federal court dismissed Elon Musk's claims against OpenAI and senior executives on May 18, 2026, after a nine-person advisory jury found that Musk waited too long to bring breach-of-charitable-trust and unjust-enrichment claims over OpenAI's shift toward a for-profit structure. The verdict followed a three-week trial in Oakland before U.S. District Judge Yvonne Gonzalez Rogers, and Ars Technica reported that Musk planned to appeal after the judge affirmed the jury's decision.

See also: Musk v Altman et al.

---

White House delays AI executive order after safety-testing dispute

The White House delayed a planned May 2026 artificial-intelligence executive order after several major technology executives declined to attend a signing event and after administration officials received objections that the order could be framed as a barrier to innovation. Ars Technica reported that the delayed order would have emphasized voluntary model-sharing with the government and AI safety testing rather than mandatory pre-release approval. Gizmodo reported that the draft order contemplated a voluntary process for companies to share frontier models with federal reviewers and that the signing event was called off after pushback from technology-sector allies.

The episode is significant for AI law because it illustrates the continuing tension between federal AI safety review, voluntary industry cooperation, and deregulatory arguments that government review could slow model deployment. It also shows that the administration's AI policy process remained unsettled even when the contemplated measure was framed as voluntary rather than as a binding licensing or approval regime.

See also: Federal Regulation, Executive Branch