GoKawiil
Tech News
Cua released CUA-S1, a specialized model for computer-use decisions, alongside Cua Driver, a toolset letting AI agents inspect and control native desktop apps and browsers on macOS, Windows, and Linux via CLI, MCP, or SDKs. The company also offers Cua Fleets, cloud-hosted sandbox desktops provisioned through run.cua.ai, so developers can test agents on isolated Linux environments without local hardware.
Newly unsealed court documents in The New York Times' lawsuit against OpenAI and Microsoft reveal internal statements warning that their AI models were creating a 'doom loop' harming the web. A Microsoft applied science director reportedly described the companies' data scraping practices as the 'largest theft of labor in human history' and a mockery of fair use. Microsoft has distanced itself from these remarks, calling them one employee's personal opinion rather than the company's official stance.
As alarm grows over AI-enabled bioweapons—fueled by Anthropic's report on misuse attempts of Claude and Stanford research on AI-designed viral genomes—some scientists argue the danger is overstated. David Bellamy of the Institute of Foundation Models says AI is not a fundamentally new threat, noting that the internet, translation tools, and open science journals already made dangerous biological information accessible long before large language models existed.
Newly unsealed court documents in the New York Times' copyright lawsuit against OpenAI and Microsoft reveal internal admissions from both companies' executives. Microsoft's director of applied science Brent Hecht reportedly described AI training as the 'largest theft of labor in human history,' while OpenAI cofounder Greg Brockman acknowledged that generative AI products pose an 'existential threat' to publishers because they are increasingly substitutive of original content.
A new open-source tool called typesafe-computer-use lets a Mac execute plain-English goals by reading the screen state directly rather than sending screenshots to a large multimodal model. It uses a lightweight classifier from TypeSafe to pick the next action from a short list of options, only invoking a text-generation model when free-form input is actually required. The developers report costs near 1/50th of a cent per decision step, versus tens of cents per step for a comparable Claude Opus-based screenshot approach.
Researchers publishing in Frontiers in Psychology observed that adult chimpanzees, mostly mothers, model tool-use behaviors like extracting hard-to-reach food, then pass their tools directly to young chimps so they can imitate the motions. The team calls this deliberate handoff a form of teaching, distinct from mere observation, that helps juveniles master foraging technology.
Cloudflare launched a 'Disallow AI Training' control that lets website owners stay visible in search results while blocking the same crawler from using their content to train AI models. Apple, Google and Microsoft have agreed to honor the new setting, which addresses the problem of mixed-use crawlers that previously forced sites to accept both search indexing and AI training together or neither.
A 2025 OECD PISA study of over 760,000 15-year-olds across 91 countries found that, after controlling for socioeconomic status, students who never use AI generally outperformed AI users in science. The performance gap was largest for those using AI to summarize texts or draft assignments, and for those using it either rarely or daily, while moderate, intentional use—especially paired with training to critically evaluate AI outputs—showed smaller drops or even slight gains over non-users.
As payments begin from Anthropic's $1.5 billion settlement over pirated training data, some authors say publishers and literary agents are wrongly claiming their share of the $3,000-per-book payouts. Writers including April Henry report publishers claiming full payments or rights to books whose rights had reverted to the author years earlier. Watchdog Victoria Strauss says the pattern likely stems from sloppy recordkeeping rather than deliberate wrongdoing, and some publishers have already asked Anthropic to correct the errors.
AI companies have started purchasing used physical books in bulk to feed data into their large language models. This practice follows a copyright settlement earlier this year in which courts determined that buying books and destroying them during the digitization process for AI training qualifies as fair use.
At a Zurich ceremony on September 3, researchers received Ig Nobel awards for offbeat but rigorous studies, including a biomechanics team that produced a cross-species definition of kissing excluding food-sharing behavior, a physics group that engineered urinal designs to minimize splashback, and soil scientists who buried and later exhumed 1,000 pairs of underwear to study decomposition rates linked to land use.
The federal government filed a letter in court supporting OpenAI and Microsoft against the New York Times' 2023 copyright lawsuit over AI training data. The filing argues that using copyrighted articles to train large language models qualifies as fair use because it is 'extraordinarily transformative' and doesn't directly compete with the Times' journalism.