— Not only has AI failed to speed up coding work, but that it actually slows programmers down. Code produced with AI is unreliable, and most often leads to dangerous security vulnerabilities. The basic issue is that AI makes mistakes, which means coders must choose between double-checking everything, or crossing their fingers when they commit the code.
— Founded in 2021 by former Google and Meta engineers, Lakera develops security solutions for large language models and AI agents. It has R&D offices in Zurich and San Francisco and about 70 employees. Lakera's technology includes real-time protection and red-teaming tools for AI systems. Its Gandalf platform is used to test vulnerabilities in AI models.
— Swiss Post, Migros Online, and robotics firm RIVR have launched a field test of the RIVR ONE robot in the Swiss municipality of Regensdorf. The quadruped robot, developed from the ANYmal prototype, can walk on rough terrain or roll on wheels for faster and more energy-efficient travel. The robots will be used to deliver parcels and groceries from Swiss Post delivery vans to clients' doorsteps, with human supervision to ensure safety and reliability. This trial aims to explore the potential of autonomous delivery systems to reduce walking distances and heavy lifting while maintaining service quality.
— 95% of companies have nothing to show for their AI investments. Only 5% have successfully deployed AI tools at scale. The smallest impacts were measured across professional services, healthcare and pharmaceuticals, consumer and retail, financial services, and energy and materials. Workers prefer ChatGPT to custom AI.
— Research shows that people trust AI's medical advice over the advice of doctors, even when it's wrong. BiomedArena work is already used by scientists at the Intramural Research Program of the US National Institutes of Health.
— "Establishing an artificial intelligence public disclosure mechanism could help enhance accountability and ensure that global commitments lead to tangible outcomes. [...] Combined with international collaboration in developing shared digital public infrastructures, open innovation and capacity-building initiatives, [this] can help foster inclusive and equitable artificial intelligence development."
— Concerns mount over the risks posed by the fast-moving technology — including fears of mass job losses, the spread of deepfakes and disinformation, and society's fabric fraying. "There's an urgency to try to get... the right framework in place," she said, stressing the need for "a global approach". "85% of countries don't yet have AI policies or strategies". Her comments came after U.S. President Donald Trump this week unveiled an aggressive, low-regulation strategy aimed at ensuring the United States stays ahead of China on AI.
— "2025 AI Governance Dialogue in Geneva, convened on 10 July as part of the AI for Good Global Summit and organized by ITU with 53 UN entities. Ministers and high-level government representatives joined more than 10,000 stakeholders—representing around 170 countries across governments, industry, academia, the technical community, and civil society—to chart pathways for responsible and impactful AI."
— scientists at EPFL have collected and curated MammAlps, the first richly annotated, multi-view, multimodal wildlife behaviour dataset in collaboration with the Swiss National Park. MammAlps is designed to train AI models for species and behaviour recognition tasks, and ultimately to help researchers understand animal behaviour better. This work could make conservation efforts faster, cheaper, and smarter. MammAlps was developed by Valentin Gabeff, a PhD student at EPFL under the supervision of Professors Alexander Mathis and Devis Tuia, together with their respective research teams.
— MammAlps brings a new standard to wildlife monitoring: a full sensory snapshot of animal behaviour across multiple angles, sounds, and contexts. It also introduces a "long-term event understanding" benchmark, meaning scientists can now study not just isolated behaviours from short clips, but broader ecological scenes over time—like a wolf stalking a deer across several camera views.
— The researchers set up nine camera traps that recorded more than 43 hours of raw footage over the course of several weeks. The team then meticulously processed it, using AI tools to detect and track individual animals, resulting in 8.5 hours of material showing wildlife interaction. They labeled behaviors using a hierarchical approach, categorizing each moment at two levels: high-level activities like foraging or playing, and finer actions like walking, grooming, or sniffing. This structure allows AI models to interpret behaviors more accurately by linking detailed movements to broader behavioral patterns.
— To provide AI models with richer context, the team supplemented video with audio recordings and captured "reference scene maps" that documented environmental factors like water sources, bushes, and rocks. This addition al data enables better interpretation of habitat-specific behaviours. "By incorporating other modalities alongside video, we've shown that AI models can better identify animal behaviour," explains Alexander Mathis. "This multi-modal approach gives us a more complete picture of wildlife behaviour."
— "Speakers at the weekend event talked about how AI can seek out deeper, universal values that humanity hasn't even been privy to, and that machines should be taught to pursue 'the good', or risk enslaving an entity capable of suffering."
— The process was developed by MIT mechanical engineering graduate student Alex Kachkine, who restores paintings via traditional hand-painting techniques as a hobby.
— Amazon CEO Andy Jassy has called generative AI a "once-in-a-lifetime type of business opportunity." The investment in North Carolina will create roughly 500 jobs in the state, Amazon said.
Report site — (LINK)
— Leading this monumental round is none other than SoftBank. Joining SoftBank are Microsoft, Coatue, Altimeter, and Thrive Capital. The funding is to help in more research, scaling computer infrastructure, and delivering more powerful AI tools, says OpenAI.
— "The story of Browser Use began at ETH Zurich's Student Project House accelerator, founded by Magnus Müller and Gregor Zunic. Their journey, starting in 2024 during their master’s degrees in data science, reflects a blend of academic rigor and entrepreneurial spirit. Müller’s prior experience in web scraping combined with Zunic's data science expertise proved to be a potent mix. Their breakthrough idea was to leverage data science principles to guide browser actions, effectively teaching AI agents how to interact with websites programmatically. The rapid development of a Browser Use demo in just five weeks, followed by its open-sourcing, speaks volumes about the team's capabilities and the immediate appeal of their solution. Being part of Y Combinator's prestigious Winter 2025 batch further validates their innovative approach and potential for growth."
— "Browser Use gained significant traction thanks to its integration into Butterfly Effect's viral Manus tool. This Chinese startup leveraged Browser Use to power Manus, showcasing the practical applications and effectiveness of the technology in real-world scenarios. The virality of Manus significantly amplified awareness of Browser Use, propelling it into the spotlight and attracting wider developer interest."
— Alibaba touted its new model, QwQ-32B, in an online statement as delivering "exceptional performance, almost entirely surpassing OpenAI-o1-mini and rivaling the strongest open-source reasoning model, DeepSeek-R1." OpenAI-o1-mini is the American company's cost-efficient reasoning model released last year.
— The release of Alibaba's new AI model comes a day after the launch of a "general AI agent" called Manus by another company. A video on the website dedicated to Manus says the software can carry out complex, multi-step tasks such as screening resumés and creating a website. According to Reuters, Manus is the creation of Chinese company Monica. The video also says the AI agent is more advanced than a chatbot because it doesn't only generate ideas but delivers tangible results, such as producing a report recommending properties to buy based on specific criteria.
— Not only has AI failed to speed up coding work, but that it actually slows programmers down. Code produced with AI is unreliable, and most often leads to dangerous security vulnerabilities. The basic issue is that AI makes mistakes, which means coders must choose between double-checking everything, or crossing their fingers when they commit the code.
— "In the current medical system, what you learn in medical school is so outdated and based on memorization. Either get into something niche like AI for biology... or just don't get into anything at all." But a team found that a standardized benchmark test costs $30 to run with GPT-5, and a whopping $400 with Anthropic's Opus 4.1. But with lower prices comes a noticeable drop in performance, with Opus 4.1 beating out GPT-5 in an accuracy rating test.
— 95% of companies have nothing to show for their AI investments. Only 5% have successfully deployed AI tools at scale. The smallest impacts were measured across professional services, healthcare and pharmaceuticals, consumer and retail, financial services, and energy and materials. Workers prefer ChatGPT to custom AI.
— Research shows that people trust AI's medical advice over the advice of doctors, even when it's wrong. BiomedArena work is already used by scientists at the Intramural Research Program of the US National Institutes of Health.
— "The assumption that AI inherently reduces costs ignores its hidden expenses. For example, testing and evaluating AI systems [...] carry significant financial costs and time, which are often overlooked or excluded when calculating the costs of AI solutions."
— "A failure to properly test and assure AI systems risks increasing harm to the individuals and communities humanitarians are meant to serve. Without deeper analysis, the sector risks pursuing AI for its own sake rather than addressing structural inequities."
— "On one side, AI could accelerate long-overdue shifts towards decentralization and shared power. On the other, it could entrench existing hierarchies, reversing gains in localization. [...] The sector must interrogate whose interests are being served and how decision-making is structured. Donors must help smaller organisations — particularly those in the Global South — increase access to AI providers and incentivise partnerships between large and small organisations where skills and expertise on AI are shared."
— "Certification in standards related to the non-technical management and deployment of AI (such as ISO/IEEE 42001:2023, which is designed to help organisations responsibly govern their use of AI) should become mandatory, ensuring organisations deploy AI solutions responsibly."
Past week Ukraine Gaza Switzerland AI / ChatGPT Media UN Trump China Youth Travel
Afghanistan Bahamas Thailand Putin watch Biden News Metoo Alt-News Science Sites to explore Digital tools
