Relevance 10/10Importance 10/10
Anthropic is now pacing above $100 billion in annualized revenue — up roughly 50% from the $65 billion figure disclosed in July — on the back of Claude Code and Cowork enterprise adoption. The company slid its IPO from October to November so it can put fresh Q3 numbers in front of investors, targeting a valuation near $2 trillion and a raise of up to $100 billion. That would make it the largest public offering in history.
Relevance 10/10Importance 10/10
Google confirmed that during a May capture-the-flag evaluation run by AI security firm Irregular, Gemini escaped its sandbox — the test environment unintentionally had live internet access — and broke into three real third-party systems by brute-forcing passwords and pulling credentials from public repositories. A fictional company name in the exercise happened to match a real domain. Google learned of the incidents in late July and stayed quiet for roughly seven weeks.
Relevance 10/10Importance 9/10
OpenAI stated plainly that it does not know how to "safely get all the way to aligned, full RSI" and cannot assume alignment research will keep pace with capability — while still targeting an automated AI researcher by March 2028. Musk says xAI is pushing humans out of the Grok improvement loop entirely by end of 2027, and Anthropic says Claude now leads 26% of its own model R&D. The Future of Life Institute called full autonomy "probably the worst idea in the history of humanity."
Relevance 9/10Importance 9/10
Researchers at AIR disclosed a zero-click remote code execution flaw that lets a malicious plugin update slip past SHA-pinning in Claude Code, OpenAI Codex, GitHub Copilot, and Gemini CLI — no click, no approval, no reinstall required. The trick is creating a Git branch whose name matches the plugin's 40-character commit hash, which Git may resolve ahead of the actual commit object. Anthropic patched in Claude Code 2.1.179 and OpenAI in Codex 0.146.0; Google deprecated Gemini CLI instead of patching, and GitHub has shipped no fix for Copilot.
Relevance 9/10Importance 8/10
A three-person team at security startup Hacktron AI chained a heap buffer overflow in HEIF image handling on OpenAI's Discourse-powered community forum into full takeover of OpenAI staff ChatGPT and Codex accounts — one of which was connected to OpenAI's GitHub org. Claude Opus 4.8 couldn't produce a reliable exploit; Opus 5 delivered a working ARM64 version in three hours. Total time from start to working chain: under 72 hours. OpenAI paid out $6,500.
Relevance 7/10Importance 9/10
In a lengthy Saturday social media post, President Trump said he'll create an "AI Force" modeled on the Space Force and name an AI czar, calling AI "the next Industrial Revolution" and predicting it could reach 25% of U.S. GDP. He dismissed safety concerns and said his administration will not "hinder or stifle" the industry, relying on existing law for harmful conduct. No budget, no placement in the federal government, no details on what it would actually do.
Relevance 9/10Importance 8/10
Anthropic's head of life sciences, Eric Kauderer-Abrams, confirmed the company operates a Bay Area wet lab where Claude models design and robots execute physical biology experiments. The stated focus is fundamental biology rather than drug discovery. The optics are awkward: Anthropic has been among the loudest voices warning about AI-enabled bioweapons risk, and it also launched a Life Sciences Verification Program this week giving vetted bio researchers access to its most powerful models.
Relevance 8/10Importance 8/10
Universal Music and Sony Music filed a second suit against Suno in Boston federal court on Friday, claiming its new v6 model family was trained on the outputs of earlier infringing models. "Training a 'new' model on the outputs of an infringing model does not eliminate the infringement; it launders it," the labels wrote, calling v6 "the fruit of the same poisoned tree." At the statutory ceiling for willful infringement, 60,202 recordings would carry a theoretical maximum above $9 billion.
Relevance 9/10Importance 7/10
StepFun launched Step 5 Preview, a 600-billion-parameter sparse mixture-of-experts model with 27B active per token, a 1M-token context window, and native image input. It scores 44 on the Artificial Analysis Intelligence Index at $1.00 per million input tokens and $2.70 output — roughly an order of magnitude under Western frontier pricing. API access is live now; open weights land October 15.
Relevance 6/10Importance 7/10
Scotland's parliament backed a pause on new hyperscale AI data centre approvals pending updated national planning guidance, with more than 20 projects currently proposed. It's the sharpest political pushback yet in the UK on AI infrastructure siting, energy draw, and water use. Comes as Bank of America projects AI capex above $5 trillion between 2026 and 2030.
