Microsoft Executive Calls AI Scraping Historic Theft in Unsealed Lawsuit Documents

According to reporting by the minority faction editorial team, a morning briefing published on September 20, 2026 summarizes several technology trends. At its core are documents unsealed in the copyright infringement lawsuit brought by The New York Times and others against Microsoft and OpenAI. They detail remarks by company executives, shifts in search traffic, and how training data was obtained. The relationship between AI training and the business foundations of news organizations is being tested in court.

Internal Documents Reveal Copyright Alarm

According to Ars Technica, Brent Hecht, director of applied science, had raised the alarm internally. He described the large-scale collection of news content as astonishing theft on an unprecedented scale. He further warned it could become the largest theft of the fruits of labor in human history.

“The largest theft of the fruits of labor in human history”

He said the principle of fair use in copyright law had become hollow. The documents show both defendants recognized that conversational AI would become a direct substitute for news outlets. Concerns about a vicious cycle destroying the supply chain were also shared internally. It is clear that the legitimacy of training and its business impact were not treated separately inside the company.

Sharp Drop in Search Traffic and

Paywall-Circumvention Allegations

Internal materials show that after the release of AI-based platforms, click-through rates from search plunged for some plaintiff outlets. The declines were recorded as ranging from 51% to 94%. The possibility has emerged that summarization features reduced the need to visit the original sites.

Communications records show an engineer reporting a way to bypass The New York Times paywall. OpenAI co-founder and president Greg Brockman replied with a like. Plaintiffs argue that news text was reproduced verbatim in response to requests for summaries or commentary. They contend readers lost the incentive to open the source.

Questions were also raised about how data was obtained. Microsoft allegedly repurposed a dataset purchased for Bing for OpenAI training. OpenAI allegedly used a third-party dataset containing 1.8 million articles despite recognizing it violated noncommercial terms. Plaintiffs’ attorneys argue the new evidence confirms awareness of wrongdoing.

Defendants Push Back as Claims Clash

A Microsoft spokesperson responded that use in its AI products constitutes transformative use. The company said Hecht’s remarks were the personal views of an employee, not its legal position. CEO Satya Nadella testified that shifts in traffic reflected changes in how information is consumed. OpenAI has not commented at this time.

Plaintiffs reject the fair use defense. They argue mass reproduction creates substitution and undermines the value of the licensing market. The defendants emphasize transformative nature and social utility. The focus will be on where courts draw the line between substitution and transformation.

Trump Proposes Renaming as Regulation Debate

Intensifies

According to TechCrunch, U.S. President Trump proposed a name change in a post on September 19. He argued that artificial intelligence is neither accurate nor fitting. He offered three alternatives — exceptional intelligence, extreme intelligence, and supreme intelligence — and asked users to choose.

He characterized safety concerns and opposition to data-processing infrastructure as political falsehoods. He stressed he would not stand by while the industry stalls and would fully support its growth. He said the planned White House AI chief must be a person of extremely high intelligence. He also mentioned creating an AI force, but did not specify its duties.

Behind this lies a tug-of-war between development speed and regulation. An Anthropic researcher resigned citing existential threats, reigniting debate over controls. In various regions, backlash over power grids and environmental impact has intensified, and New York State has suspended permits for a large infrastructure project. Meanwhile, Nvidia CEO Jensen Huang opposed slowing development and expressed support for the post.

Android 17 Pixel Exclusivity Sparks Ripples

According to GrapheneOS, Android 17 QPR1 released September 15 includes Pixel-only APIs. It says this is the first release to introduce new APIs without syncing them to AOSP. It describes the shift as the biggest since Android 3.x. Other manufacturers and the open community are not expected to gain access until QPR2, scheduled for December 2026.

Starting with Android 16, Google changed its release policy for quarterly versions. It no longer pushes QPR1 and QPR3 to AOSP, publicly releasing only annual major versions and QPR2. The September Pixel-only update included fixes for the generic platform. However, without syncing to security bulletins for all manufacturers, fixes are reportedly delayed on non-Pixel devices. Delays in providing source code for the GPL kernel were also criticized.

GrapheneOS says it has completed support for QPR1 but must maintain it through porting low-level drivers until public release is permitted. Citing growing difficulty in third-party maintenance of Pixel devices, the group is pursuing an official partnership with Motorola. Balancing open development with prioritizing its own devices has become a supply-chain-wide challenge.

Battery Limits and Case Against Screen

Protectors Explained

Apple published a support document titled Getting Ready to Ship Your iPhone 18 Pro Max. The device’s battery capacity exceeds 20Wh. In many countries and regions, devices with a single cell exceeding 20Wh face additional shipping requirements. From factory shipment through transport, battery-related programs limit capacity to under 20Wh. The company explains the limit is lifted after activation.

The company’s vice president of device technology said screen protectors are not recommended. The reason is believed to be the display’s durability and tactile design. The impact on the accessories market is drawing attention. Adjustments to operations and specifications are underway on both shipping requirements and hardware design.

China State Railway Group on September 19 urged purchases through official channels. Ahead of Mid-Autumn Festival and National Day travel starting September 23, it continues to detect malicious hoarding. It identifies bulk purchases and device spoofing with high accuracy, diverting them to slow queues and refusing ticket issuance. Since the effort began, 7.117 million cases have been diverted to slow queues, ticket issuance was refused in 1.023 million cases, and 1.331 million tickets were denied. It warned that third-party channels do not improve success rates and risk delays or failures.

Editorial Opinion

In the short term, release of the lawsuit documents is seen tightening scrutiny of training data collection by AI developers. Licensing negotiations with news organizations are likely to accelerate, making increased cost burdens unavoidable. With figures corroborating declines in search traffic now disclosed, the bargaining power of media outlets is assessed to strengthen.

In the long term, judicial rulings on the scope of transformative use are seen shaping the design of training infrastructure. Building only with licensed data and strengthening source attribution could become standardized. For users, designs that make the origin of information clear are assessed to matter more than response speed.

The remaining question may be who decides the line for substitution. How should declines in page views be weighed against the benefits of providing summaries. Just as with user-burdening specification changes such as screen protectors and battery limits, the question is who should bear the cost of information use.

References

• “派早报：微软高管称 AI 爬取是人类历史上最大的劳动成果盗窃”, by 少数派编辑部 — 少数派, 2026-09-20T22:26:04.000Z (ARR)
• Source URL: https://sspai.com/post/114788