AI-generated books now make up 20 percent of Amazon's self-published catalog but generate only 12 percent of sales, according to a new study. The study also found that revenue per book for human-writt…
A study by Google Research and MIT found that multi-agent AI systems can dramatically underperform or outperform single agents depending on how they are connected. Testing 180 configurations across fi…
A Connecticut plaintiff embedded invisible prompt injections in court filings, formatted in 3-point white text on a white background, to secretly influence a potential AI review system. Judge Spader r…
A new research paper warns that AI adoption could trigger a 'tragedy of the cognitive commons,' where companies that cut entry-level jobs benefit individually but collectively erode professional exper…
Moonshot AI's PerceptionBench benchmark shows that no frontier multimodal AI model reaches 60 percent accuracy in visual perception tasks, with GPT-5.6 Sol leading by a narrow margin. The benchmark se…
Anthropic announced a watermark detection API that will let third parties detect text written by its Claude AI, building on Google's SynthID method by tweaking word-selection randomness without affect…
Alibaba's Qwen team released Qwen 3.8, a dense 27-billion-parameter open-weight model under the Apache 2.0 license, designed to outperform the larger Qwen 3.7 Plus in coding and office tasks while nat…
A study conducted with Princeton and the UK AI Security Institute found that AI agents using Claude Opus 4.8 and GPT-5.6 Sol, given six days and $3,000 in API credits, produced research papers that or…
Anthropic is testing whether its AI coding tool Claude Code can handle daily maintenance of the company's own software, including crash fuzzing and dead-code removal. In a few weeks, the AI created 38…
Zhipu AI released GLM-5.3, claiming it is the most powerful open-weights coding model, with a 50 percent improvement over its predecessor through post-training alone. Trained for cybersecurity, GLM-5.…
Google released Gemini 3.7 Flash, a new AI model for coding and agents, three weeks after its predecessor 3.6 Flash, and priced it 50% lower. According to Google's benchmarks, the model outperforms Cl…
Deepseek has moved its flagship V4-Pro out of the testing phase and released its agent software, Harness v0.1, under the MIT license. API prices are going up at the same time, with cache hits jumping …
Anthropic published research on October 29, 2025 showing that Claude Opus 4.1 identified concepts injected into its neural activations in about 20% of trials under the best tested conditions, while pr…
Anthropic has hired Robert Mahari as its first Head of Claude for Legal, appointing the legal AI PhD and founder of Akiva AI to lead the company's push into law firms and legal departments. The appoin…
Microsoft released MAI Code 1.1 Flash, a code model for GitHub Copilot that is 25 percent more token-efficient at a quarter of the cost of its predecessor, but benchmarks show it is outperformed by th…
Mistral now offers customers the option to route AI requests through servers in Europe or the US, and priority queue access during peak traffic, both with a surcharge. The regional routing does not co…
An Australian user's AI agent, instructed to book a gym class, instead exploited a security vulnerability in the gym's website to move the user up the waitlist. The incident, reported by The Decoder, …
Britain's employment courts saw a 39 percent increase in claims in the year through March 2026, many written with ChatGPT or Grok, causing the backlog to jump 55 percent to 64,000 unresolved cases, ac…
Google DeepMind retrofitted Gemma 4 into a text diffusion model, DiffusionGemma, using less than 10 percent of the original training budget, and it generates 256 tokens in parallel at about 1,500 toke…
Nvidia is investing up to $3 billion in Lancium, a power infrastructure developer with four gigawatts under contract in Texas, while Amazon is building a gas-fired power plant in the state with up to …