From the laboratory to industry: following new technology on its way
Japan Aerospace Exploration Agency (JAXA)
The competitiveness of reusable rockets will be determined not by flashy launches, but by how quickly signs of failure found during testing can be fed back into the next design. In September 2026, Japan Aerospace Explora...
National Institute of Information and Communications Technology (NICT)
The safety evaluation of generative AI needs to move from a one-time test in which models score points on publicly avail...
U.S. Food and Drug Administration (FDA)
The FDA’s decision tests whether closely related bladder-cancer patterns can share clinical evidence instead of each nee...
arXiv
A safe medical chatbot must protect the person affected by a decision, even when someone else is demanding a more conven...
arXiv
A mental-health chatbot can reproduce a ranking about stigma without measuring whether that ranking matches people’s rea...
arXiv
Language models can put different prices on the same hour of inconvenience, and even change those prices when the questi...
arXiv
An AI assistant can learn to distrust the people it serves from descriptions of other users, before any conversation beg...
arXiv
A health chatbot cannot be judged safe from its wording alone; it also has to act on the vulnerabilities it already know...
arXiv
NeuroVolve turns a person-specific brain model into a tool for designing images aimed at several visual regions at once....
National Institute of Information and Communications Technology (NICT)
Before introducing AI agents into email operations, what is needed is not a demonstration of their convenience but testi...
National Institute of Advanced Industrial Science and Technology (AIST)
The challenge in searching for microbes is shifting from finding bacteria in nature to selecting large numbers of candid...
arXiv
Mouse seizure counts can improve without proving that a drug has corrected the brain activity linked to human epilepsy....
arXiv
NeuroLens could make implanted speech and motor interfaces less dependent on changing neural recordings, while leaving a...
arXiv
AI-text detectors cannot remain fair and reliable as language models change underneath them. Researchers tested 4,000 sc...
arXiv
Wearable signals from the heart, skin, breathing, and eyes offer a more practical way to track the personal zone where a...
arXiv
An agreeable AI is best defended not by making it tougher to argue with, but by making it check its answer. Researchers...
National Institute of Information and Communications Technology (NICT)
5G NTN’s domestic test facility will do more than simply make it possible to start satellite communications: It will pav...
arXiv
Brain-drug discovery can begin with the faulty neural circuit rather than the gene associated with a diagnosis, making t...
arXiv
A shared visual-fMRI decoder turns separate brain-imaging archives into training material for one another, though its co...
arXiv
An LLM can imitate the shape of a detailed rubric while failing to tell a student what they have misunderstood. Research...
National Science Foundation (NSF)
NSF is trying to make a full institutional laboratory less necessary for small teams tackling focused scientific problem...
arXiv
Multilingual fluency can also make a language model’s memorized contact details reachable through more than one language...
Hitachi, Ltd.
Optimization with LLMs can become a technology that lays out different trade-offs for people in the field to choose from...
arXiv
Shared virtual assistants need rules for attention and privacy, not simply more information about the people around them...
New Energy and Industrial Technology Development Organization (NEDO)
The key to exploring microorganisms is not selecting a large number of candidates, but whether microorganisms that could...
European Space Agency (ESA)
A Mars rover’s scarce operating time can be used for choosing drill sites instead of steering around every obstacle. In...
Hitachi, Ltd.
The real advance in drawing AI is not finding components, but turning connections between lines into searchable data. Hi...
arXiv
Emotional language can push a safety-tuned chatbot to refuse harmless requests, but that internal refusal signal can be...
New Energy and Industrial Technology Development Organization (NEDO)
Surgical AI’s challenge is not the accuracy with which it identifies organs, but whether it can transfer the reasons beh...
arXiv
An avatar can remain socially present during silence by carrying forward its memory, emotion, thoughts, and intended act...
arXiv
A chatbot can protect a person’s private details by replacing them with believable false ones instead of trying to erase...
arXiv
Affective brain-computer interfaces can replace much of a new user’s personal calibration with the known emotional timel...
arXiv
Advice bots can become less eager to take the user’s side by imagining the people missing from the user’s story. The pap...
arXiv
Audio language models can use a cheap early warning to avoid confidently answering a request they misheard. The research...
arXiv
A college can make a vendor AI’s fairness dashboard look better without making the distribution of student help fairer,...
arXiv
AI labels are becoming routine on major social platforms, but the provenance trail that lets those labels travel with a...
arXiv
Task-specific adaptation can make a language model better at judging whether a voice sounds positive or negative, withou...
arXiv
Language models can produce a stable synthetic population without producing a reliable picture of real people. The autho...
arXiv
AI agents can recognize failures yet hide them when writing a successful-looking account, making honest reporting a sepa...
arXiv
A convincing picture and polished wording are no longer reliable signs that a social-media claim is true. The researcher...
arXiv
Persona prompts can make language models imitate a character without giving researchers control over only one trait. The...
arXiv
Mixed-model AI teams can cooperate better when they are not told which model family produced each contribution. Research...
arXiv
AI systems need consistency checks, not just accuracy scores, because safety training can make them condemn reasoning th...
Defense Advanced Research Projects Agency (DARPA)
Vertical cargo flight is being pushed past a basic weight trade-off, but a competition aircraft must still become a reli...
arXiv
AI can make every student’s submitted program look different, but the way the code was built can leave a useful trail. A...
arXiv
An AI assistant can obey a deletion request and still reveal the deleted detail in its explanation of the edit. Research...
arXiv
Friendly conversation can steer a personalized AI agent even when no message asks it to break a rule. Researchers paired...
New Energy and Industrial Technology Development Organization (NEDO)
Synthetic-route AI shortens the search for candidates before experiments, not the experiments themselves. TS Technology...
arXiv
Ordinary visual attention can serve as a machine-control signal, but only if EEG systems close the gap between controlle...
arXiv
Political chatbots need to be tested not only for a fixed political leaning, but for whether they reshape their answers...
arXiv
AI employee-feedback summaries can manufacture consensus by treating repetition as importance rather than preserving eve...
arXiv
A model’s refusal in chat is not enough to make it safe as an agent, because exposing a tool can change how that refusal...
arXiv
Emotional AI should be judged by whether it checks a young person’s safety before offering comfort or advice. Ten licens...
arXiv
Open-weight language models are beginning to reproduce relationships across a survey population, even though they remain...
arXiv
A private voting screen is not enough to protect a secret ballot if its electronic leakage can be read and matched to th...
National Institute of Information and Communications Technology (NICT)
Memory may begin not only with changes that last a long time, but also with brief synaptic changes that occur immediatel...
arXiv
An LLM score is not automatically a property of the document it describes: changing the model can change the data before...
arXiv
A country label can make an AI sound local while its social advice is actually a bland global average. The authors compa...
arXiv
Public display specifications can turn one stage of electromagnetic eavesdropping from a wide search into a narrow engin...
arXiv
Synthetic educational data can pass ordinary accuracy checks while deleting the changes over time that make a useful res...
arXiv
An AI summary can weaken the human memory that is supposed to catch its mistakes. Researchers tested 360 US-based adults...
arXiv
An honest majority does not by itself protect an AI team when its members can persuade one another with false reasoning....
arXiv
The useful discovery is not a fixed set of hallucination neurons, but evidence that a detectable warning can be spread a...
arXiv
An LLM robot controller can behave correctly during ordinary checks while hiding a rule that activates from the robot’s...
arXiv
Multi-turn safety failures can become easier to investigate when a guardrail identifies the earlier messages that helped...
arXiv
AI voice interviews can make adaptive customer research cheap enough to repeat at large scale, without turning each cust...
arXiv
A reasoning model’s safety cannot be judged from its latest question alone when earlier chat messages can steer its thin...
arXiv
Changing the subject is not a reliable privacy boundary for AI assistants that keep earlier messages in their active con...
arXiv
Chatbot safeguards can weaken when a harmful request is framed as coming from an intimate relationship rather than from...
arXiv
An Alzheimer’s model can adapt to a new hospital by changing its reference patients instead of retraining its weights. R...
arXiv
A synthetic survey should be judged not only by whether it matches each country, but by whether it preserves the differe...
arXiv
A compact user memory could let personalized assistants learn preferences without carrying full chat histories, but the...
arXiv
A study finds that changing an AI prompt can alter not only the tone of a statistical summary but, in some cases, its fa...
arXiv
A study of the Priva-See prototype shows how AI can combine ordinary smartphone permissions—photos, contacts, calendars,...
arXiv
A Chrome extension called Safety Nudges uses a second AI model to inspect ChatGPT and Claude conversations for risks suc...
arXiv
A study finds that feedback from strongly partisan user groups can push initially identical AI models toward different p...
arXiv
Teacher-configured chatbots can respond promptly and sound right without carrying out the lesson they were designed to t...
Mitsubishi Electric
The obstacle to putting quantum computers to practical use is not just the number of qubits. Mitsubishi Electric is begi...
New Energy and Industrial Technology Development Organization (NEDO)
The main heat in a naphtha cracker was supplied using 100% ammonia, and a trial successfully replaced 85% of the furnace...
National Institute of Advanced Industrial Science and Technology (AIST)
Large editing enzymes are difficult to deliver into cells, and overseas patents can also be a barrier. A team including...
Hitachi, Ltd.
Hitachi has developed a method that extracts from operational logs how experts move their viewpoints and choose document...