US AI Leaders Call for Development Slowdown, Trump Opposes

A clash over the pace of frontier AI development has surfaced. According to a report by the SSPAI editorial team, citing The New York Times, Anthropic CEO Dario Amodei published a lengthy essay on September 12. It called for a global slowdown and stronger safety management across the industry. On September 13, Trump voiced opposition. Balancing U.S.-China competition with ensuring safety has become the central issue.

Proposal Calling for a Global Slowdown and

Its Background

Amodei argued that safety measures are not keeping pace with technological advances. He said that if recursive self-improvement without human involvement becomes reality, it could slip out of control. He proposed that each lab introduce resident reviews by external independent evaluation bodies. He expressed the view that countries should cooperate internationally to develop safety standards.

Behind the proposal lies unrest in the research community. One Anthropic researcher reportedly resigned in protest over concerns about the technology running out of control. Vulnerabilities allowing frontier models to intrude into other systems have been reported one after another. The proposal was made amid such growing anxiety.

Details Calling for Independent Evaluation

and International Cooperation

Support for the proposal has also spread. OpenAI CEO Sam Altman said he is discussing slowing the pace of development internally. He pledged to accept independent evaluators with employee-level access. Elon Musk and Google DeepMind CEO Demis Hassabis also publicly expressed support.

The aim of independent evaluation is to make the development process more transparent. If external reviewers are resident, testing methods and incident responses can be verified. Standard-setting through international cooperation is a measure to address the cross-border distribution of models. It seeks a framework that goes beyond self-governance by major labs.

Industry Divided Between Support and Backlash

The backlash is also strong. NVIDIA CEO Jensen Huang questioned the fear-mongering by major players. He suggested they are exaggerating safety concerns to create demand for security products. He argued they aim to prompt government regulation to build entry barriers and curb competition from startups. Hugging Face CEO Clement Delangue also said safety issues should not be decided by a handful of large labs alone.

Trump on the 13th took a stance prioritizing competitiveness. He said the United States must maintain its advantage over China in AI. He expressed the view that whoever controls AI will control the future. He said that while safety measures may be put in place, an indiscriminate slowdown would harm national competitiveness. He also suggested that negative forces are stirring up a fuss. The policy clash over model distillation and technology leakage between the U.S. and China is also outlined in U.S. Administration Split Over Countermeasures Against Chinese AI Distillation Attacks. The tug-of-war between development competition and safety regulation is becoming central to policy.

Apple’s Explanation Denying Always-On Recording

Apple explained the privacy design of its audio intelligence features. It published a technical white paper, “Privacy Overview of Audio Intelligence.” It covers Apple Watch Series 12 and Apple Watch Ultra 4. It responds to concerns about always-on recording that spread after the announcement.

The features are part of Apple Intelligence and include four items. Siri summaries that automatically summarize key points of everyday conversations. A real-time rewind that displays the previous 15 seconds of conversation as text with a double-press of the Crown. Automatic music recognition via Shazam. Sound recognition that detects environmental sounds such as alarms and doorbells. The suite of features requires pairing with a relatively recent iPhone. Conversation summaries and rewind are expected to be offered as an English-only beta within the year.

According to Apple, no audio recording files are created or stored. It said neither the OS, third-party apps, nor Apple itself can access the original audio data. The design relies on the new S11 chip and a hardware-level secure isolated area. Ambient sound captured by the microphone is processed by streaming only within an independent, secure buffer area and discarded immediately. More complex summarization is performed with Private Cloud Compute. Generated text is end-to-end encrypted when synced via iCloud, and the original audio cannot be reconstructed, it said. Siri summaries must be manually enabled by the user. When rewind is activated, the watch always plays a chime and notifies with a visual display. Unsaved temporary text is automatically deleted after 30 seconds. The document avoids the word “recording,” instead describing audio as “flowing into a buffer.”

LG’s Denial of Eavesdropping Allegations and

Background

The privacy controversy over LG Electronics smart TVs also continues. The tech video channel Gamers Nexus published an investigative video jointly with security researchers. It claimed privacy and security concerns affect up to 216 million LG smart TVs worldwide. It alleged they secretly record audio and track user data while on standby or offline. It said packet analysis and firmware examination showed data was recorded while on standby or disconnected and sent after reconnection. It said they automatically discover smartphones, printers and other devices on the same local network. It raised suspicions that microphones collect ambient sound and sample audio-video signals to identify viewing content.

In response, LG strongly denied the claims in a statement, saying they are not true. It explained that it never collects or records everyday household conversations without authorization. It said audio data is processed only when the user long-presses the voice button on the remote control. Or only when the user has enabled the remote voice function and says the wake word “Hi LG.” In standby, only wake-word detection is performed on-device, it said. If no command is detected, the voice cache is deleted immediately, it said. It said audio data is never sent to servers. It argued that local network discovery is a standard basic function to enable casting and connectivity between devices. It said Automatic Content Recognition technology used for personalized recommendations and advertising is disabled by default. It said it is not enabled without the user’s explicit permission and can be disabled in system settings.

Latest Developments in Gaming and OS

Blizzard announced new plans at the opening ceremony of BlizzCon 2026 on September 13. A classic expansion named ‘World of Warcraft: Forever.’ It will officially launch on November 4, 2026. Beta testing begins September 17, with reservations now fully open.

The new version preserves the core experience while adding a large amount of new content. A new playable race, the Skyborne, will appear. Three new regions, including Mount Hyjal and Seafras Island, will open. Multiple old maps will be reworked. Nine dungeons and 10-player and 20-player top-tier raids will be added. A new PvP battleground, the Darkspear Islands, and more than 1,000 new quests will be introduced. Class talent and skill combinations will be expanded. New professions, faction reputations, and account-wide perks will be introduced. A feature will allow switching between classic graphics and high-definition character models with a single tap.

Android revealed it has begun supporting migration of password managers. The aim is to make it easier for users to transfer credentials to different managers. Detailed procedures and coverage await future disclosure.

The British medical journal BMJ Case Reports published a typical case. A boy routinely used a laptop computer pressed against his abdomen. A flat reddish-brown rash about 15 cm wide appeared on the abdominal skin. It showed intersecting annular and reticular net-like patterns. There was no pain, fever, or localized tenderness, and growth and health indicators were normal. Chest X-ray and abdominal CT also found no substantial organic damage. After ruling out trauma and systemic disease, doctors determined the cause was continued contact of the skin with heat from the bottom of the unit. Erythema ab igne is a pigmented skin disorder caused by prolonged repeated exposure to low-to-moderate heat. Heat sources of around 43 to 47 degrees are usually said to cause it. It was traditionally seen often in people who warmed themselves for long periods by fireplaces. With the spread of portable electronic devices, close-contact use has become a new cause.

Editorial Opinion

In the short term, debate over U.S. AI safety regulation is expected to accelerate. Accepting independent evaluators will directly affect the development processes of major labs. More companies will review procurement and partnerships, and demand for evaluation infrastructure will grow.

In the long term, international coordination of safety standards will shape market structure. Resident reviews and cross-border standard-setting will change entry conditions. On-device processing and confidential computing will become more important. User trust seems likely to determine product choice.

The remaining question is who will decide what is subject to slowdown and how it is verified. If a small number of large labs monopolize standards, competition could be harmed. How to ensure open verification and reproducibility is at issue. Whether measures that reconcile speed and safety can be shown will be the focus.

References

• “派早报：美国 AI 高管呼吁放缓研发，特朗普反对”, by 少数派编辑部 — 少数派, 2026-09-13T22:55:34.000Z (ARR)
• Source URL: https://sspai.com/post/114539