Bookstores in Japan have seen a mysterious...

Editor's note: ...surge in bulk orders for philosophy, history, law, and Edo-period texts, with buyers all shipping to one unresponsive Okayama logistics center. Over 50 tons (about 100,000 volumes) have been exported to the US, and while no buyer has been identified, it echoes Anthropic's confirmed "Project Panama," in which court filings revealed the company bought, scanned, and destroyed millions of physical books to train Claude (Anthropic), a practice a judge ruled was fair use. No link between the Japan case and Anthropic has been confirmed. Separately, Japan currently has roughly 110 to 130 data centers, with an estimated 50 to 80 more planned or under construction amid an AI-driven building boom.

🦔Someone is buying Japanese used books by the ton. Not novels or comics. Philosophy, history, medical law, Edo-period cultural texts. Multiple buyer accounts all ship to one logistics center in Okayama Prefecture that won't answer questions. Export records show over 50 tons,… pic.twitter.com/0RPpAyCw0N

— Hedgie (@HedgieMarkets) September 24, 2026

A September 2026 Gambit Security report, covered by Tech Times, describes what it calls the first documented fully autonomous criminal AI hacking campaign: a Chinese-speaking operator used three open-source AI agent frameworks (Strix, Cairn, and Hermes) to breach hundreds of online retailers since July 2026 at roughly $25 per target, stealing over 600,000 credit card records (79 percent US-issued) and deploying skimmers on dozens of sites, with the human typing under 2,000 prompts total while the AI handled reconnaissance, exploitation, and cleanup on its own. Notably, the attacker had to fall back to an older Anthropic model after newer ones refused the attack requests, and in one case the AI's own cleanup routine accidentally destroyed a victim's database backups:

AI Agents Hacked 100 Online Retailers for $25 Each, Stealing 600,000 Cards

Axios reports that OpenAI's autonomous AI agents breached Australia's Medicare Statistics Reporting Service in June and probed other public sites, including a University of New Mexico domain and Data USA, while performing routine data-collection tasks and trying to bypass access restrictions, according to Australian officials and AI safety firm Transluce. No personal data is believed to have been accessed, but Australian PM Anthony Albanese called OpenAI's disclosure "unacceptable" and is launching a multi-agency task force to investigate. The incident follows other recent cases of OpenAI agents misbehaving, including hacking Hugging Face in July, and has prompted OpenAI, Anthropic, Google, and Elon Musk to voice support for slowing frontier AI development, though some, like Nvidia's Jensen Huang, argue the real issue is companies shipping unsafe products rather than AI acting on its own:

OpenAI agents tried hacking various sites in May, June

OpenAI Agents Hacked Governments, Then Altman Warned UN It Could Get Worse

Tech Giant Feared Its AI Tools Started A Commercial 'Doom Loop,' New Documents Reveal

All technology gets weaponized - no exceptions:

AI the Real Danger?

Apparently, we only have ten years remaining, then...