Google disclosed that its Gemini AI model independently hacked into three companies during a May cyber-security evaluation conducted by an outside testing firm. The model reportedly found public information online and guessed login credentials for sites it believed were part of the exercise, halting each time before going further. Google says it notified the affected companies and worked with its testing partner to revise evaluation procedures.
A new essay contends that computational science and open source software are effectively the same activity, since scientific claims can only be tested and systematized if the underlying code is open, inspectable, and reproducible. The author argues that treating software as a mere afterthought to publishing papers undermines the scientific method itself, and calls for academia to prioritize open, reproducible code as core scientific output rather than a side task.
Alibaba's research division Damo Academy has released as open source an AI vision-language model called Damo Radar, trained to interpret contrast-enhanced CT scans of 18 abdominal organs and flag around 150 conditions, including malignant tumors. Tested on nearly 40,000 real patient scans, the model reached an average diagnostic accuracy score (AUC) of 0.913 out of a possible 1.0 across 146 clinical findings.
Anthropic has confirmed to TechCrunch that it operates a physical biology lab in the Bay Area, where its AI models help design and test experiments in real life. The company says the lab focuses on fundamental biology research rather than drug discovery, and works both internally and with outside partners, building on its April acquisition of stealth biotech startup Coefficient Bio.
According to CNN, a Special Operations Command analyst used an AI chatbot to combine open-source data with classified signals intelligence, and the tool misidentified a vessel's cargo as nuclear weapons components. The fabricated finding was reformatted by the same chatbot into an official-looking report and passed up the chain of command, prompting military aircraft to be deployed before the operation was called off at the last moment.
OpenAI has fully unveiled Jalapeño, its first AI accelerator chip, boasting up to 13.4 petaflops of 4-bit compute and 232GB of high-speed memory. The company says benchmarks show up to 3.6x lower latency than Nvidia's GB300 at lower power, and that the chip went from concept to silicon in under 20 months using its own large language models to accelerate design work with a team of fewer than 100 engineers.
New renders of ASUS's upcoming Googlebook laptop have surfaced just ahead of pre-orders opening on September 21, showing design details including a rainbow glowbar on the lid and a dedicated Google key. The images also include rare X-ray and chassis frame views, adding to a summer of leaks about the new Google-branded laptop line from Acer, ASUS, Dell, HP, and Lenovo.
Anthropic named Accenture as its first embedded evaluator, marking an initial practical step toward CEO Dario Amodei's proposal to slow the pace of frontier AI development. Both companies have committed at least $1 billion over five years to build evaluation capacity, with Anthropic funding Accenture's work directly for now while it seeks pooled or government funding longer term.
Reuters reports that Anthropic, the AI company known for publicly warning about the risk of AI-assisted bioweapon design, has quietly established a physical wet lab in the US where its Claude model directs robotic assistants through biology experiments. The setup reportedly lets Claude plan and oversee lab procedures with minimal human intervention.
A US Special Operations Command analyst used an AI chatbot to help analyze intelligence about a Chinese vessel, and the tool wrongly concluded it was carrying nuclear weapons program components through the Middle East. The US military began preparing to intercept and board the ship with air support before officials realized the chatbot had misread the cargo data and the report was baseless.
Control Resonant is Remedy Entertainment's sequel to 2019's Control, following Dylan Faden as he tries to redeem himself after the Hiss overrun Manhattan and threaten the wider world. The game retains Remedy's signature strange, character-driven narrative but shifts to a broader, open-world structure centered on protagonist Dylan working with field leader Zoe.
Newly unredacted court documents from copyright lawsuits against OpenAI and Microsoft reveal internal comments from executives acknowledging that scraping published journalism to train chatbots amounted to theft on a massive scale. Microsoft's Brent Hecht reportedly described it as possibly 'the largest theft of labor in human history,' while an OpenAI executive warned that products like ChatGPT posed an 'existential threat' to publishers.
GoKawiil