---

🧠 I. Large AI Models

Flock Camera Controversy Heats Up Midterm Elections, Anti-AI Sentiment Becomes Campaign Issue

Axios reports that privacy controversies surrounding public safety camera operator Flock continue to ferment, with calls for boycotts and paused partnerships appearing across the US. Flock's license plate recognition network has spread into many communities nationwide. How long data is retained and how it is shared with police have always been focal points of resident protests. Criticism isn't limited to communities; anti-AI sentiment has been directly dragged into the 2026 midterm elections, with AI applications in law enforcement becoming an unavoidable topic for candidates for the first time.

Questions about the scope of data collection are driving more local regulations. Cities are already reassessing contracts with Flock. For companies like Flock in the public safety AI space, pressure is mounting from slowing new signings and contract re-evaluations. For the broader legal tech sector, this controversy serves as a lesson for all AI apps targeting police. The battlefield for AI policy debate has expanded from Washington to state and local levels, amplifying corporate political exposure risks during election cycles. The subsequent developments of this event are worth tracking, and discussions around commercializing public data will heat up accordingly.

The landing point of anti-AI sentiment is shifting from tech blogs to state legislatures and campaign rallies, meaning compliance costs for AI companies are no longer just engineering issues but now include a political variable. For AI companies operating in the US, those with a higher proportion of public sector clients need to build buffers into their data governance early. The financing environment for legal tech will also sway with legislative trends in key states, with tightening probabilities looking higher currently. The evolution speed of the Flock incident warrants special tracking, as the intensity of US AI regulation can be gauged from it. This case reminds the market that the AI industry has truly entered an election cycle for the first time. Policy disturbances will appear as frequently as tariff issues. Investors need to place political cycles and industrial cycles on the same timeline, considering both in budgets and valuations. If federal AI legislation continues to stall, fragmented state-level actions will persistently drive up compliance costs—a variable to factor into models for the coming quarters. Leading companies have already begun bolstering their public affairs teams.

---

• OpenAI Launches Teen Version of ChatGPT — Multiple child safety experts question the transparency of its safety mechanisms. Minor protection becomes a new battleground for AI products. Whether product grading can earn regulatory trust depends on the disclosure level of safety design.

💻 II. AI Software

US Companies Begin Catching Up on AI Trust Lessons; Layoff Fear is Biggest Obstacle

A CNBC survey shows that the more prevalent AI becomes, the higher employee anxiety rises. The most common pitfall enterprises encounter when deploying AI is skipping communication and going straight to system implementation. Employees use the tools while worrying about job displacement. Now, many companies are starting to include rollout pacing and role impact explanations in deployment plans. Trust management has become a mandatory part of AI projects.

Management sees efficiency gains; employees see changes to their livelihoods. This cognitive gap is becoming the most underestimated cost in enterprise AI adoption. The survey also found that companies that have experienced layoffs face significantly greater resistance in subsequent AI projects. Grassroots resistance to tools translates directly into efficiency losses. Companies must make "communicate clearly before implementing" a fixed process, which tests organizational capability more than model selection. AI governance and change management capabilities are quietly entering enterprise software selection checklists.

This topic benefits two types of roles: service providers standardizing AI implementation consulting, and software vendors embedding change management tools into their products. Employee trust crises won't disappear in the short term, but every failure case helps newcomers calibrate deployment pacing. Mature companies are solidifying these experiences into processes; internalizing them into product features is just a matter of time.

• Linkdaze Releases Home Edition AI Calendar — This product doesn't just record schedules; it hands off household chores like grocery shopping, pickups, and bill payments to AI coordination. From morning prep to evening cleanup, family arrangements are integrated into one dynamic timetable. The AI assistant moves from the office desk to the living room, targeting the scenario of managing a household. AI applications in home scenarios have lacked entry points; high-frequency tools like calendars might be a smoother starting point than smart speakers.

The difference from big-company office calendars is that it treats family members, not colleagues, as collaboration objects. Privacy boundaries and data ownership will determine how far such products can go. Sensitivity of family data is far higher than work schedules. Who holds this data and how it is monetized will be repeatedly questioned by regulators and markets during the product design phase. Watch for follow-up moves by other manufacturers. Once validated, vertical family assistants like this are easily copied by giants. Small teams' window of opportunity won't last long; first-mover advantage must be converted into user habits early. Once migration costs for calendar data are established, competitors will pay double to steal users.

• Norway Plans Comprehensive Ban on Campus AI Glasses — The Minister of Education urges local governments to write bans into school rules. Norway previously restricted AI use in schools; this time, the ban extends to wearable devices. Following phones, AI glasses in classrooms become the new regulatory focus. Wearables are always recording and querying. How regulators measure this is being watched by other countries.

This line may also affect future commercial expectations for AI hardware entering campuses. Eyewear manufacturers targeting the education market need to incorporate compliance design into product planning early. European countries' cautious attitudes toward AI in classrooms are converging. Consumer AI hardware should note that the regulatory window for wearables is shifting from function review to scenario review. Campuses, offices, and public spaces will have separate rules. Manufacturers' scenario-based compliance capabilities will become a new competitive dimension. AI glass makers' paces are diverging; some choose to start with industrial and enterprise scenarios to avoid consumer market regulatory minefields. High-sensitivity scenarios like sports events and exam halls are likely the first places where lines will be drawn.

• Andrew Ng Updates AI Engineering Skills Map — Breaks top-tier skills into two main lines: building and deploying AI apps. Corresponding engineering capabilities, evaluation methods, and toolchains are expanded layer by layer, giving developers a clearer training path. The skills map has undergone several revisions; the biggest change this time is significantly raising the weight of deployment-side capabilities, consistent with industry talent gaps.

Industry focus is spreading from model training to both pre-engineering and post-ops. Hiring market signals match this: engineers who understand deployment, evaluation, and data pipelines are harder to recruit than pure algorithm roles. For training institutions and university courses, this map is basically an official highlight list. A batch of courses and certifications will be reorganized according to this framework. Enterprise recruiters can also use this map to compare against job descriptions, saving trial-and-error costs.

• Open Source Tool Prism Reviewer Launches on GitHub Marketplace — A multi-agent code review tool built on LangGraph and LiteLLM. It can automatically run code reviews in CI pipelines, posting review comments directly back to Pull Requests. Another member joins the AI programming collaboration ecosystem. Behind the dense appearance of such tools is the fact that code review is becoming the first aspect automated in AI coding implementation.

Unlike Copilot-era assistive completion tools, multi-agent review takes a full-process managed route. Competition between the open-source camp and closed-source vendors in this slice has just begun. Code review is the quality gate for enterprise developers. Handing it to agents requires quantifiable review standards, a premise increasingly accepted by teams. The match between review standards and agents will become a key selection metric. Worth tracking later is whether it can meet enterprise private deployment needs—that's the market with real willingness to pay.

• Wondershare Global Launch of Filmora.TV — Listed AIGC company doubles down on AI video creative track. Based on Filmora, it creates an AI canvas workspace for commercial TV ad creators, integrating infinite canvas, script, and storyboard generation. A-share AI application companies move from tool export to deep professional scenarios.
• Blogger Detained for Fabricating Company "IPO Failure" — Self-media blogger fabricated an internet company's IPO failure by piecing together information and was administratively detained. Costs and legal risks of AI-generated rumors rise simultaneously, increasing content platform moderation pressure.

---

• OpenAI Chief Global Affairs Officer Lehane Warns of AI Cyberattacks — Public and enterprises must prepare defenses against AI cyberattacks. The double-edged sword effect of AI enters the core of national security agendas. Security products and insurance markets will expand following this judgment.

🤖 III. Humanoid Robots

HK University Professors Enter Entrepreneurship En Masse, Embodied AI Startup Ecosystem Expands

QbitAI reports a wave of professor-entrepreneurs emerging from Hong Kong universities. They publish papers at top conferences and are founders or chief scientists of billion-dollar valuation companies. Interviewed professors describe their state as having one foot in the lab and one in the factory, turning academic results directly into product prototypes. After Unitree's listing, the heat in the embodied AI primary market transmitted to academia. Hong Kong is becoming a new startup hub.

Professor entrepreneurship raises the technical threshold for early-stage projects and compresses the time from paper to prototype. Industrialization pace is visibly accelerating. Teams with HK university backgrounds have become new scarce assets in the primary market. Valuation games around these entrepreneurs have begun. OEMs feel that massive influx of academic power means increased mobility of talent and technology. The competitive landscape may change faster than imagined, and the combination methods of research institutes and industrial capital will be redefined.

Hong Kong's role is worth observing separately. Backed by the Greater Bay Area supply chain and connected to global capital and academic networks, these teams have natural channels for both financing and mass production. Technical competition in embodied AI is shifting from hardware specs to algorithms and data accumulation. Professor teams' accumulation in data collection and simulation methods is their most undervalued asset. Short-term, orders from HK university teams mainly come from research institutes and high-end manufacturing clients. Long-term, it depends on whether technical accumulation can convert to bulk shipments. Hong Kong's continuous investment in innovation resources provides policy soil for these teams. R&D collaboration networks within the Bay Area are their most ready-made support for the mass production stage.

• 2027 Beijing Yizhuang Humanoid Robot Half-Marathon Opens Global Invitations — Event incentives and assessment standards upgrade simultaneously, opening registration to global research teams. Unlike regular marathons, humanoid robot contestants must pass both technical assessments and track tests. Stability, battery life, and gait algorithms are exposed during the race.

Another validation window added to the humanoid robot arena. Beijing Yizhuang's intent to normalize the event is clear. Participating teams iterate annually; race results are basically public versions of each company's technical progress bars. From last year's inaugural event to the dense competitions around this year's World Robot Conference, the humanoid robot evaluation system is being standardized by events. Impact on the supply chain goes beyond race day; training, supplies, and venue sensors will settle into reusable technical facilities. The reference value of this window will rise yearly, facilitating investors' horizontal comparison of actual levels of various OEM solutions. Which metrics will be written into assessments next year? Everyone will study this in advance. The race...