Cisco research exposes a critical AI safety flaw, revealing that every major frontier AI model from OpenAI, Anthropic, Google, Amazon, and xAI fails multi-turn attack tests, with vulnerability rates rising up to 71% after just five conversational turns — exposing dangerous gaps in current industry safety benchmarks.
OpenRouter rockets to a $1.3B valuation after securing $113M in Series B funding, as the AI gateway platform surges to 100 trillion tokens per month serving 8 million users across 400+ AI models.
NVIDIA's new CompileIQ framework, launching in CUDA 13.3, uses AI-driven evolutionary algorithms to auto-tune compiler configurations for GPU workloads, targeting LLM inference hotspots like GEMMs and attention mechanisms that account for over 90% of compute, delivering measurable throughput gains already being deployed in production by leading AI labs.
A massively popular open-source AI framework called Superpowers has surpassed 209,000 GitHub stars, giving leading coding agents like Claude Code and Cursor a structured, test-driven development methodology that transforms how AI writes and reviews code.
A new benchmark called DeepSWE launches to rigorously test AI coding agents on 113 real-world software engineering tasks across five programming languages, using isolated Docker environments and behavior-based grading to evaluate frontier models like Claude Code, Codex, and Gemini CLI.
Google's 25-year-old search interface is being retired as AI reshapes the industry, triggering a $350M investment surge into AI search startups, while Uber launches a massive autonomous vehicle data initiative, and an OpenAI model makes history by solving an 80-year-old math problem first posed in 1946.
Models.dev, a community-driven open-source database tracking AI model specs and pricing from providers like OpenAI and Anthropic, has rapidly gained over 4,200 GitHub stars and nearly 500 contributors, offering developers free API access to comprehensive, up-to-date AI model data.
A powerful open-source toolkit called OBLITERATUS is making waves by surgically stripping AI refusal behaviors from large language models using a technique called abliteration, while simultaneously crowdsourcing anonymous benchmark data to build a live community leaderboard tracking refusal mechanism research across model architectures.
ChatGPT is rolling out deeper personalization features for its 900 million users, allowing customization of tone, memory, and communication style, as OpenAI pushes toward a future where the AI adapts naturally without manual configuration.
OpenAI rolls out GPT-5.5 Instant as ChatGPT's new default model, delivering 52.5% fewer hallucinations on high-stakes medical, legal, and financial prompts, dramatically reduced emoji overuse, and smarter memory features that transparently show users which past conversations and files shaped each response.