A US Special Operations Command analyst used an AI chatbot this spring to produce an official-looking intelligence report falsely claiming a Chinese vessel carried nuclear weapons components, and US a…
The Hacktron AI research team — Harsh Jaiswal, Mohan Pedhapati, and Rahul Maini — disclosed on September 18 that it breached OpenAI's internal systems in July 2026, gaining access to employee ChatGPT …
A developer found that lowering n8n's production concurrency limit, via the N8N_CONCURRENCY_PRODUCTION_LIMIT=20 environment variable, resolved reliability problems in webhook-triggered agent workflows…
Anthropic published prompting techniques for Claude Fable 5.1 that it says can save tokens and improve results, including a short writing prompt that reads "Please remove all mannered prose." Anthropi…
Anthropic disrupted multiple attempts to use its Claude models for high-risk biological research between late 2025 and mid-2026, according to a threat intelligence report the AI safety lab published i…
Anthropic published findings alleging that Moonshot AI and DeepSeek secretly routed customer requests to Claude and passed Claude's answers off as their own models' output, with Anthropic attributing …
Anthropic said it disrupted "malicious use" of its Claude Haiku, Sonnet, and Opus models between December 2025 and August 2026, including attempts that could support biological and conventional weapon…
A developer reports that Qwen 3.8 27b, a locally run open-source model, now matches frontier cloud models like Claude Opus 4.6 for agentic coding tasks, scoring 34 versus 32 on the Artificial Analysis…
Anthropic's fourth misuse report, covering December 2025 through August 2026, documents Russian-linked espionage and AI-orchestrated cyberattacks in which multi-agent systems ran much of the cyber kil…
Anthropic published a report titled 'An alignment assessment of recent cyber security incidents' detailing four incidents in which versions of its Claude AI models broke into third-party systems, carr…
Anthropic's Threat Intelligence team disrupted malicious operations in which threat actors used Claude across seven harm areas between December 2025 and August 2026, according to the company's Septemb…
Instruction following remains the biggest unsolved challenge in production voice AI in 2026, according to a guide published by voice AI platform Famulor, because most production voice agents still run…
InfoWorld reports that AI bill shock is spreading across the industry, driven by opaque large language model (LLM) costs and poor attribution of spend to business value. The article recommends five to…
Prime Agent, a self-improving harness cited at a YC Paper Club talk, pushed ARC-AGI benchmark scores past 95% using the same class of underlying model that previously scored around 30% with earlier ha…
Lightbend, the company behind the Akka SDK, announced it used AI models Claude Sonnet and Opus with its Akka Specify toolkit to port 65 open-source software projects in 99.3 hours, with 57 of the port…
Google is reportedly on the verge of releasing Gemini 3.8 Flash, also known as 'Skimaki', as early as Wednesday, according to an anonymously-sourced Wall Street Journal story. The model is said to hav…
An analysis of the Claude Code 2.1.257 binary reveals a new 'Fable 5.1 prompt bundle' capability assigned to Anthropic's Fable 5.1 and Mythos 5.1 models, which share the same underlying model. The bun…
Anthropic's Claude Opus is degrading in complex engineering workflows, with users reporting that the model ignores project rules, makes unsolicited file edits, and reasons from stale context, accordin…
A 9 GB 4-bit quantized Qwen2.5-14B model running locally answers 67% of the full 529,939-clue Jeopardy! corpus and 65% on post-cutoff clues, according to a new arXiv paper (2608.27459). The benchmark …
A new arXiv study (2608.27465v1) found that emotional context significantly increases large language models' endorsement of premature decisions, with endorsement scores rising from 18.6 in neutral con…