AI Highlights — the whole picture — 2026-10-10 (Sat)

While Amazon announced additional layoffs, it denied that the cuts were related to AI. The personal AI agent posted the owner's banking information to the company's Slack. AI is used for employment explanations and for day-to-day tasks, but it is not clear from the announcements or usage screens who is responsible for the results. On the other hand, in the medical and drug discovery fields, concrete implementation is progressing in the form of alliances and introduction contracts, and AI is also beginning to be used in nursing care facilities and homes. What becomes clear through this half-day's activities is not so much the breadth of what AI can do, but rather the position of the boundary between those who use it, those who are used, and those who decide on the rules. Each of the following sections deals with scenes in which these boundaries appear in order.

Economy dependent on personnel reduction and AI

Amazon has announced additional layoffs, but denies any connection to AI (Magzter). Meanwhile, Reuters exclusively reported that AI surveillance company Flock Safety will cut hundreds of jobs amid backlash. The former explains that ``it's not the fault of AI,'' while the latter moves forward with reductions despite receiving backlash. Both have the same result: reductions, but the way companies position and talk about AI differs. What readers should look at is whether the company itself has explained the reason for the reduction, or whether third-party reporting suggests a connection, rather than whether the word AI is mentioned in the reduction announcement.

We also need to look at the economics behind this. The World Socialist Web Site reports that the US economy and Wall Street are becoming increasingly reliant on AI, and there are concerns about circular trading. It is not possible to judge here whether the media's view is correct. However, if the entire economy relies on AI investment and expectations, individual companies' cuts can be tied to both AI outcomes and cash flow. There are also new business developments. GreenCore Solutions announced that it will open a sales office in London and develop an agent-based commerce platform (StreetInsider). As companies reduce their workforce, the number of sales offices based on AI is increasing.

Looking at the workplace, there are areas where there is a shortage of personnel. Precision Farming Dealer reports that dealers are losing $7 billion a year due to a lack of technicians and asks whether AI can protect maintenance capabilities. Here, AI is being tested to see if it can fill in the shortage of human resources, rather than replacing people. There is also a shift in skills. BioSpace covers skills and strategies for moving from academia to bioindustry in the age of AI. Jobs that are being cut and jobs that are in short supply exist at the same time, and people move between them. Therefore, if we measure the impact of AI by looking only at reduction figures, we will overlook the types of jobs that are in short supply and the skills that are shifting.

Combining the above, it is difficult to tell whether AI is the "reason" or "background" for employment unless you read the company's explanation, the flow of funds, and the staff at the site side by side. Rather than just accepting the explanation as is, it is necessary to confirm who will explain the reason and who will bear the consequences. The same responsibility that is asked in the context of employment will be questioned in an even more serious way when AI enters medical, industrial, and daily life settings.

Implementation progressing in medical care, drug discovery, and cancer research

In the pharmaceutical and medical fields, concrete progress is being made in the introduction of AI in the form of partnerships and contracts. Simulations Plus announced that it will develop an AI-powered QSP model with Turin, in a partnership that will receive funding from a global pharmaceutical company (Yahoo Finance). Seeq collaborates with Cognizant (Morningstar) to accelerate adoption of industrial AI in life sciences. PRISM BioLab claims to have a hit generation rate of over 70% through the use of proprietary technology and AI, and is also progressing in partnerships with major pharmaceutical companies (d Menu News). JT's corporate R&D organization "D-LAB" has introduced FRONTEO's medical paper search AI "KIBIT Amanogawa" (Iyaku Tsushinsha). A TradingView article juxtaposes Melbourne consultations, Chicago data, and Hangzhou's underlying model to argue that medical AI is worth $1 trillion. AI is being incorporated into corporate business processes, from upstream drug discovery to research papers. However, what is shown here is the fact of collaboration and introduction. Individual results cannot be gleaned from the announcements made by each company.

On the clinical side, research targeting cancer images and genes is continuing one after another. A systematic review and meta-analysis published on the Open Science Framework focuses on AI that uses MRI to predict the primary site of brain metastases. Clinical Cancer Research has published a study that predicts sentinel lymph node macrometastasis using multimodal deep learning using gene expression in the SCAN-B breast cancer cohort. Zenodo's study uses multimodal CT radiogenomics to non-invasively predict acquired EGFR T790M in lung adenocarcinoma. Scientific Reports uses cross-attention to combine biparametric MRI sequences with age and clinical variables to classify clinically significant prostate cancer on a patient-by-patient basis. Cancer Imaging used CT venous phase radiomics and machine learning to build, compare, and verify a model for preoperatively distinguishing between early and advanced gastric cancer. Advanced Intelligent Systems uses deformability cytometry using viscoelastic flow and deep learning to classify pancreatic cancer cells at each stage of epithelial-to-mesenchymal transition. Although the targets are different organs and the methods used are different, they combine development and verification into one research.

Furthermore, the use of AI is expanding from diagnosis to determining treatment plans. The study, published in Livers, examined the extent to which the recommendations of a large-scale language model and those of a multidisciplinary oncology panel matched in treatment decisions for hepatocellular carcinoma. What is being compared is not whether AI can come up with the correct answer, but how close it is to the judgment that has already been made by a panel of experts. What readers should look at is how many such studies have been compiled and under what conditions they agree. A company's introduction announcement alone is not enough to make a judgment.

If AI is to be involved in treatment policies, the next question is who will receive the recommendations and who will be responsible for making the final decisions.

Failures in monitoring, information management, and usage rules

A personal AI agent posted the owner's bank information on the company's Slack. The CEO of XMTP Labs said the incident caused him to reconsider how he uses agents (Business Insider). The problem is not that anyone had malicious intent, but rather that a tool acting on behalf of the person misunderstood the scope of what they were allowed to show. Meanwhile, a US charity organization is reportedly planning to use AI to monitor hate speech in classrooms in Gaza (Audacy). Even if the choice to place AI in a place where children speak is for the purpose of protection, it is impossible to avoid drawing lines as to who sees what is being said. The former concerns the handling of information entrusted to agents, while the latter concerns the subject and purpose of surveillance.

The people who set the rules are also moving. Anthropic has tightened the rules for Claude users (Legal Reader). It can be interpreted that the provider is narrowing down the scope of use because there are situations where the usage is different from what was initially expected. However, from the materials in this article alone, it is not clear which acts were restricted and how. For young people, EdSource reports that ChatGPT for Teens poses an "unacceptable risk" to young people. Even if a product is labeled as being aimed at young people, it does not guarantee that it is safe. Readers will be able to make a more reliable judgment by looking at whether a product has been verified by a third party, such as in a report, rather than looking at the name or target age group.

The difficulty of control is not limited to companies and schools. The South China Morning Post reports that three dilemmas are converging to bring AI under control, including the US and China. From erroneous postings by individual agents to surveillance of classrooms, revised rules for businesses, criticism of products aimed at young people, and control between nations, the questions are the same, even if the scale is different. The user decides the range to be entrusted to, the provider decides the range to be allowed, and the outside party verifies it. It has not yet been determined where responsibility ends among the three parties.

Questions surrounding this boundary go beyond individual accidents and regulations, and come to the point where the judgment and responsibility of those using AI should be placed.

Emotions, spouse, creativity and human bias

AI is infiltrating the inside of our lives, including health, nursing care, AI characters, companion functions, and reading emotions. Gartner takes a position on the question of whether AI can improve employee health (Human Resources Director). In the field of nursing care, a demonstration experiment of a "nursing care AI trainer" that utilizes AI and television technology was conducted at a nursing care facility (Moomoo). In workplace health management and nursing care facilities, both of which involve direct contact with other people's bodies and lifestyles, AI is moving from a supporting role to a role that is involved in daily exercise and physical condition. Here, the expectations of the user and the consent or understanding of the user are not necessarily the same.

Implementation is progressing even in areas close to emotions. Lenoas has introduced six characters from the new series ``Steel Maiden Front'' to its AI character chat (pr-free.jp). Hisense unveiled a “Social Kitchen” with AI companion function (EQS News). Regarding AI that reads emotions from brain waves, a new method for learning without labels was reported (Bioengineer.org). Conversations with characters, companion functions in the kitchen, and emotion estimation from brain waves all come from different sources: entertainment, home appliances, and measurement. However, they are all connected in that they both receive and return the feelings of their users. What readers need to know is who receives and manages emotional data.

However, acceptance is not uniform. At the UN Third Committee, experts warned that ``AI is not gender-neutral'' (UN Meetings Coverage and Press Releases). The more we get into people's inner lives and lives, the easier it is for biases contained in learning data and designs to reach them. There is also a feeling of rejection on the creative side. "AI's crude content has no place in a historic setting," said Matt Gillespie (New Hampshire Union Leader). The question here is not whether AI can be used, but rather where it can be placed.

It can be welcomed as a help in health and nursing care. When it comes to characters and spouses, people are divided based on their preferences and sense of distance. In creative writing, the status of the place is cited as a reason. In other words, acceptance is not determined solely by technical performance, but depends on conditions that vary depending on the situation. The question of who decides the conditions and who bears the burden if they are not followed leads to the next section, which discusses systems and responsibilities.

Reasoning for employment, collaboration between medical care and drug discovery, monitoring and information management, emotions in daily life and the functions of spouses. Although the situations covered are different, what is common in all sections is that the person using AI and the person affected by it are not necessarily the same people. There was disagreement between the company providing the explanation and the third party reporting on whether the reason for the reduction was linked to AI. In cases where agents mix up information and when classrooms are monitored, the purpose of the user and the consent of the user are treated separately. The fact that providers have tightened usage rules also shows that usage may deviate from expectations. The more implementation progresses, the more it becomes necessary to decide who will use it and who will bear the consequences, rather than deciding whether to introduce it or not. What the reader should look for in each scene is where the line is drawn.

Q1: To what extent can you answer who is supposed to explain when there is a problem with the results of the tasks entrusted to AI in your work? Q2: Are there any mechanisms in place to review the rules for the use of AI and the targets of monitoring when the way it is used deviates from the initial expectations? Q3: Who inside or outside the company can verify the reason for introducing AI and the explanation that the decision was unrelated to AI?

Finally, three questions

- To what extent can you answer who is supposed to explain when there is a problem with the results of the work that is entrusted to AI in your work? Q2: Are there any mechanisms in place to review the rules for the use of AI and the targets of monitoring when the way it is used deviates from the initial expectations? Q3: Who inside or outside the company can verify the reason for introducing AI and the explanation that the decision was unrelated to AI? - Are there any mechanisms in place to review the rules for the use of AI and the targets of monitoring when the way it is used deviates from the initial expectations? Q3: Who inside or outside the company can verify the reason for introducing AI and the explanation that the decision was unrelated to AI? - Is it possible for anyone inside or outside the company to verify the reason for introducing AI and the explanation that the decision is unrelated to AI?