💸 Anthropic reports $80.7 billion loss ahead of IPO

🔒 ChatGPT now lets you lock chats behind a PIN

🪖 US Army builds AI agents to hack enemy networks

🧬 OpenAI buys Orchestra Bio to expand biotech AI

🚗 Google brings personalized Gemini to Android Auto

+ 📊 Daily poll and results

+ 📈 Trending tools and resources

Anthropic posted an $80.7 billion net loss in the first half of 2026, up from $13 billion a year earlier, according to a draft IPO prospectus that Reuters and The Wall Street Journal reviewed.

A $72 billion accounting charge drove most of it, reflecting the rising value of its convertible bonds. Cash costs grew too, with compute spending hitting nearly $9 billion, up from $2.3 billion.

Some investors now weigh a valuation near $1.5 trillion, below earlier $2 trillion talk. With OpenAI eyeing a listing next year, investors see this IPO as a test for the whole AI market.

OpenAI started rolling out Locked Chats, which hides chosen ChatGPT conversations behind a four-digit PIN or phone biometrics and pulls them from regular history and search, across plans from Free to Enterprise.

Locked conversations move to a Locked chats folder under Projects. New messages there won’t feed your regular memory, and OpenAI says the lock adds privacy without adding encryption or changing data retention.

One catch: resetting a forgotten PIN permanently deletes every locked chat, and OpenAI Support can’t recover them. Your secrets stay safe from roommates, and also from you.

The U.S. Army is developing ECHOTRIBBLE, a program for self-replicating AI agents that move through enemy networks, find targets, and “neutralize” them before self-destructing, according to documents The Intercept obtained.

Army Cyber Command wants agents that work like computer worms, copying themselves across Windows and Apple devices, hitting cloud and air-gapped systems, operating with minimal human oversight, and concealing transmissions to stay unnoticed.

The Army already picked one of 65 vendor bids on October 8, yet ARCYBER didn’t answer The Intercept’s questions about stopping the agents from wrecking civilian systems along the way.

OpenAI acquired Orchestra Bio, a startup whose AI agents handle the operational grind of drug development, like tracking budgets and untangling clinical trial delays. Neither company disclosed terms, but Orchestra had raised about $20 million.

Orchestra’s team and tech will join Rosalind, OpenAI’s life sciences models and tools, to build agents linking lab findings to drug development decisions. OpenAI already lists Moderna and Novo Nordisk as Rosalind partners.

Orchestra came out of stealth about 14 months ago. For startups building software for big industries, getting bought by an AI lab beats the alternative: watching that lab ship your product as a feature.

Google started rolling out Personal Intelligence to Gemini in Android Auto, letting the assistant pull from your Gmail, Calendar, Maps, Photos, YouTube, and Search. Cars with Google built-in get it in the coming months.

Gemini can grab flight details from a Gmail confirmation to check their status, or handle “take me to my doctor’s appointment.” The feature stays off unless you opt in, and you choose which apps connect.

Drivers across most of Europe and in Nigeria won’t get it yet, and neither will Workspace accounts. Your work inbox stays out of the passenger seat, at least for now.

A) Yes - 67% 🏆

B) No - 33%