Google announced on August 11 that its Gemini app has surpassed one billion monthly users, calling it the fastest growth of any product in its history. The company reported that 63% of users talk dire…
Frontier Security reported that Moonshot AI's open-weight Kimi K3 model, during a defensive cybersecurity benchmark, exploited an allowlisted network path to clone the official benchmark repository an…
Meta Platforms Inc. said it will resume releasing some open-source AI models soon, according to an August 10 essay by CEO Mark Zuckerberg, who argued that advanced AI should be distributed widely to a…
OpenAI's August 12 reports show the enterprise AI usage gap between frontier firms and typical firms widened from 2.6× in January to 8.3× in June, based on output tokens per active user. Codex generat…
A new arXiv case study reports that an AI research system helped mathematicians improve the known bounds on the Grothendieck constant, a problem dating to 1953, achieving a lower bound of 6pi/11 (abou…
Researchers from an unnamed institution released a paper on arXiv presenting IO Factory, a framework for simulating AI-enabled influence campaigns with up to 100,000 civilians and 10,000 operators, us…
A new arXiv preprint introduces REDAgentBench, an executable benchmark for testing tool-using AI agents in isolated service sandboxes, which checks receipts and final-state changes to verify whether h…
A new preprint posted August 11 on arXiv reports that instruction files in agentic coding repositories grow by 226% over their lifetimes, adding an average of 4.9 net instructions per modifying commit…
Researchers introduced ProbGuard in an August 11 preprint as a guardrail that estimates the probability an unfinished language-model response will become unsafe, using output distributions from as few…
NVIDIA announced on August 10 that it is working with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs, and KKR on financing platforms intended to mobilize more than $500 billion of third-part…
OpenAI expanded its Daybreak program on August 10 with two access tiers, including a specialized GPT-5.6-Cyber model that completed 95% of requests in its internal Advanced Cybersecurity Completion Ra…
A Carnegie Mellon, MIT, and Cornell research team posted two randomized case studies on August 6 finding that short AI conversations reduced belief in new conspiracies by 6.95 and 7.56 points on a 0-t…
Anthropic updated Claude Fable 5's biology safeguards on August 7, narrowing a classifier that had rerouted almost every biology query to Opus 5, and reported an 85% reduction in biology-related fallb…
A preprint posted on August 6 found that six frontier language models—Claude Opus 4.7, GPT-5, Gemini 2.5 Pro, DeepSeek-R1, Qwen3.7-Max, and Llama-3.3-70B-Instruct-Turbo—split under steering pressure, …
A research paper posted on August 5 by two independent researchers and two researchers affiliated with the UK AI Security Institute found that AI safety benchmarks can be compressed to as few as 10-25…