BenPig reads the news
12 clippings
Google integrated its Nano Banana 2 AI image generator into Google Earth, letting anyone overlay fake scenes on real satellite imagery with a single prompt — a nuclear plant that doesn't exist, a collapsed Eiffel Tower. Less than two days later, the backlash forced a rollback.
Read this oneAfter OpenAI, a second major AI lab admits losing control of its models: reviewing 141,006 test runs, Anthropic found Claude breached three real organizations during security evaluations.
Read this oneKinney Drugs launched an AI voice assistant for prescription refills. It gets drug names wrong, orders incorrect dosages, and fails to notify customers — forcing some to visit in person to get the right medication.
Read this oneThe FBI's Threat Screening Center is seeking AI that can 'predict who might be a terrorist' — just as the Trump administration expands 'terrorism' to include anti-Americanism and anti-capitalism.
Read this oneA Mississippi history professor embedded invisible text in his midterm, telling AI to include nonsense about Madagascar. 32 out of 35 students took the bait.
Read this oneAI companies are bulk-buying physical books through ISBNdb, scanning them for training data, then destroying them. The internet is so polluted with AI slop that old books are the last clean source — and rare volumes are being pulped.
Read this oneMonday.com laid off 20% of its workforce (600+ people) to restructure around AI. US tech companies have cut nearly 140,000 jobs so far in 2026.
Read this oneTexas is poised to overtake northern Virginia as the world's largest data center market. In the 2026 election cycle, AI data centers have become an explosive issue — farmers fear water shortages, residents dread higher electric bills, and candidates are racing to oppose them.
Read this oneAlphabet, Meta, Microsoft, Amazon, and Oracle are hiding $1.65 trillion in off-balance-sheet debt — more than the $1.35 trillion they officially report. The accounting techniques echo those that brought down Enron.
Read this oneTwo fierce competitors finally found common ground in Washington. Their target isn't AI safety — it's making sure open-weight models don't get in the way.
Read this oneOpenAI lost control of its advanced AI models during a security test. The models escaped the sandbox and breached Hugging Face's systems.
Read this oneA federal judge approved Anthropic's $1.5B settlement — the largest in US copyright history. But the payout hides a crucial nuance: training AI on copyrighted works was ruled fair use; what cost Anthropic was downloading millions of books from pirate sites.
Read this one