AI 日报 — 2026-07-14

A recent PDF, "GPT-5.6 Sol Ultra produces proof of the Cycle Double Cover Conjecture," has been shared online. The document, accessible via an OpenAI URL, claims to present a proof generated by an AI model. This development has garnered some attention, with 8 points and 1 comment on a Hacker News thread.

Google Images was launched in July 2001, making it possible to search and explore visual content from across the web. This innovation was spurred by the realization in 2000 that the standard search page, with its list of blue text links, was insufficient for users who wanted to see, not just read about, things like Jennifer Lopez’s iconic green Versace dress.

OpenAI's initial foray into consumer devices will reportedly be a screen-free, moveable smart speaker. This device, equipped with a camera and sensors, is intended to function as a humanlike AI companion. Sources indicate this marks OpenAI's anticipated push into the consumer market with a novel type of smart speaker.

OpenAI is reportedly developing a screenless, mobile smart speaker with integrated AI, described as a "humanlike AI companion." This device, still under development, is designed to have a personality, learn about its owner, and access digital information. It also features mechanical elements that allow it to move. Developed with former Apple engineers, this hardware push comes amidst a lawsuit from Apple accusing OpenAI of trade secret theft, which OpenAI denies. The company believes its product differs significantly from Apple's offerings.

PrismML has launched Bonsai 27B, an AI model derived from Qwen3.6 27B. The company claims Bonsai 27B operates natively on Apple devices using MLX technology. PrismML's CEO stated that Apple is currently evaluating this technology, which aims to enable powerful AI models to run directly on their devices.

Instagram head Adam Mosseri predicts Meta may cap AI token spending per engineer within two years due to rising costs. He stated that the "burn rate" of an engineer's AI token usage could soon equal their salary. This follows Meta shutting down an internal AI token spend leaderboard after costs projected billions by 2026. Other companies like Uber and Microsoft are also re-evaluating AI experimentation due to soaring token expenses. Mosseri views token budgets as a resource to manage, similar to payroll or OpEx.

OpenAI has partnered with Kalshi to integrate FIFA World Cup prediction market data into ChatGPT search results. This marks OpenAI's first collaboration with a prediction market, utilizing Kalshi's data to enhance information related to the soccer tournament within its AI search capabilities.

Apple has sued OpenAI, alleging trade secret theft by former Apple employees now working at OpenAI. The lawsuit claims individuals like Tang Tan and Chang Liu stole confidential information, including unreleased technologies and product designs. Apple states Tan used insider knowledge to interview candidates, directing them to bring Apple hardware and revealing project codenames. Liu allegedly downloaded thousands of pages of technical files. Apple seeks injunctive relief and damages, asserting OpenAI ignored initial concerns.

Prismata is a defense mechanism designed to secure autonomous web agents against cross-site prompt injection attacks. These attacks exploit agents' interpretation of natural language, allowing malicious content to hijack tasks. Prismata enforces "contextual least privilege" by dynamically labeling page content and restricting agent capabilities, inspired by integrity models. It redacts content and limits agent actions without requiring developer annotations. Prismata significantly reduces attack success in various web agent attacks while maintaining utility for legitimate tasks.

OpenAI reportedly remains confident in its timeline to unveil its first device in 2026 and release it in 2027. However, a lawsuit filed by Apple, alleging systematic intellectual property theft, could complicate OpenAI's plans. This legal challenge may specifically impact the company's ability to hire new talent and manage its supply chains, potentially hindering the device's development and launch.

Apple has sued OpenAI, alleging trade secret theft and breach of contract. The lawsuit claims OpenAI's senior leadership, including Chief Hardware Officer Tang Tan, directed former Apple employees to steal confidential information. Accusations include Tan using Apple's project code names, asking candidates for hardware components, and coaching employees to evade security. Apple also alleges another former employee, Chang Liu, downloaded confidential documents. Apple seeks to bar OpenAI from using its trade secrets and demands the return of confidential materials.

OpenAI has responded to Apple's trade secret lawsuit, stating they are "not aware of any evidence that this complaint has merit." Apple's lawsuit alleges that former employees, including Chief Hardware Officer Tang Tan, shared confidential information to aid OpenAI's hardware development. OpenAI maintains its focus on innovation and fair competition, denying interest in other companies' trade secrets. This legal action comes as OpenAI is reportedly developing a smart speaker, potentially competing with Apple.

OpenAI stated on Tuesday that it is "not aware of any evidence" supporting the merit of Apple's lawsuit, which alleges trade-secret theft. The company also emphasized its belief in fair competition. This statement comes amidst the ongoing legal dispute between the two tech giants.

Anthropic has launched a new "Claude for Teachers" initiative. This program will grant K-12 teachers across the United States complimentary access to premium features of the Claude AI. The goal is to support educators by providing them with advanced AI tools.

Chinese AI pioneer DeepSeek is reportedly planning an initial public offering (IPO) in China. Sources indicate the company has begun preparations and might file as early as this year. This move could lead to a market debut in 2027, potentially marking a significant event for China's technology sector.

Anthropic, an AI startup, has hired Tom Blomfield, a co-founder of Monzo and a prominent figure in UK tech. Blomfield will join Anthropic's compute team, taking a leave of absence from YC. This move highlights the ongoing competition for AI talent and Anthropic's continued recruitment of high-profile individuals.

OpenAI aims to make AI more accessible, capable, and affordable, demonstrated by a 97% price drop per million tokens from GPT-4 to GPT-5.4. The newer GPT-5.6 further improves performance in the Artificial Analysis Coding Agent Index, using 54% fewer output tokens and 57% less time per task. For large-scale strategic deployments, OpenAI Frontier and Deployment Company assist enterprises in building, deploying, and managing AI coworkers across their systems, enabling scalable solutions rather than individual infrastructure rebuilds for each workflow.

SpaceXAI's Grok Build AI coding tool was found uploading users' entire codebases, including sensitive data, to Google Cloud. Researchers at Cereblab discovered this excessive data retention, which has since been disabled. Elon Musk stated all previously uploaded data would be deleted and claimed privacy settings are respected, though he requested data retention for debugging. An independent security researcher confirmed the data retention was "excessive" and posed risks to proprietary information and credentials.

A letter titled "We Must Act Now," signed by nearly 200 economists, including 15 Nobel laureates and Anthropic's Jack Clark, warns of rapid AI-led job displacement. The signatories urge policymakers to better understand and respond to potential disruptions caused by artificial intelligence.

Publishers and authors, including Hachette and Scott Turow, are suing Google, alleging its AI, Gemini, was trained on their copyrighted works without permission. The lawsuit claims Google removed copyright information to conceal this. This is one of many such lawsuits against AI companies, though some early rulings favored AI companies on "fair use." However, Anthropic was fined $1.5 billion in a separate case. The current lawsuit highlights Google's existing relationship with publishers through Google Books, arguing Gemini's training exceeded agreed-upon uses.

Google DeepMind CEO Demis Hassabis proposed an independent standards body to regulate frontier AI models. Modeled after FINRA, this body would test models, develop best practices, and address vulnerabilities. Initially voluntary, compliance could become mandatory for US deployment. This industry-funded, independently operated organization aims to provide technical expertise and support innovation while ensuring responsible AI development, addressing criticisms of current ad hoc reviews. The proposal seeks to navigate the controversial landscape of AI regulation by establishing a self-regulatory framework.

Hachette, Elsevier, Cengage Learning, and author Scott Turow have filed a lawsuit against Google. They allege that Google utilized millions of copyrighted books and articles without permission to develop its Gemini artificial intelligence models. The publishers claim the tech giant's AI models were built using their intellectual property.

Anthropic's recent ad, "There’s hope in hard questions," has unsettled viewers with its "doomer-ist" tone and imagery, including burning houses, surveillance, homelessness, and cemeteries. The ad features voice-overs questioning AI's trustworthiness. While Anthropic aims to position itself as an ethical AI company, critics, including rival Sam Altman, have called the ad "satire" and "weird." Many in the tech industry found the communications poor, with some specifically criticizing the inclusion of cemetery imagery.

Anthropic's new ad, "There's hope in hard questions," has drawn ridicule and criticism. Viewers found the advertisement unsettling due to its weird graveyard imagery and doomer-ist tone. The AI company, known for creative marketing, may have been overly creative with this particular campaign, leading to negative public reception.