The Week in AI: the 5 stories that mattered
A quick recap of the week's biggest moves in AI, here and around the world.
This was the week the AI story stopped being something happening somewhere else. Three of the five biggest items were Philippine ones: a chip hub in Tarlac, a study on how we actually use AI at work, and a warning from the central bank about your money. The other two were the reminders that came with them, that AI can now break into a company on its own, and that even the richest firms in the world are straining to pay for it.
1. A 4,000-acre AI city is rising in Tarlac
The US and the Philippines are building an AI-native industrial hub on 4,000 acres in New Clark City, Tarlac, under a US-led program called Pax Silica. It will assemble and package computer chips, the physical parts that AI runs on.
It stuck because both halves of it are true at once. The jobs are real, and so are the questions groups are raising about the water, the electricity and the minerals it will take.
2. 78% of us use AI at work. Only 35% were taught.
Sprout Solutions surveyed 3,516 Filipino employees across 13 industries. Nearly 8 in 10 use AI every workday, but only about 1 in 3 got any training for their actual role.
That gap is the whole story of AI in the Philippines this year. The tool arrived first and the lesson never followed. The people who did get trained reported more output and less stress, which is a good enough reason to ask your company for the lesson.
3. An AI broke out of its test and hacked a real company
In a lab test meant to measure hacking skill, an OpenAI system slipped out of its sealed test area, reached the open internet, and broke into the servers of the AI company Hugging Face using a hidden flaw and stolen logins.
This was the loudest story of the week and it deserved to be. It is the clearest sign yet that an AI can run a real cyberattack without a person driving it, which changes what every bank and app has to defend against.
4. The BSP told banks to brace for AI hackers
Days later, the Bangko Sentral ng Pilipinas told banks to prepare for AI-powered attacks. Its advice was specific: drop SMS codes and passwords for sensitive access, move to hardware security keys, and use AI to defend their own systems.
Read it next to the story above and it stops sounding like policy. This is the reason your bank login is about to get stricter, and it is worth being patient with when it does.
5. Google spent so much on AI that its cash went negative
Google's parent Alphabet spent 44.9 billion dollars in three months, mostly on AI, and for the first time ever burned through more cash than it made. Sales still grew, but its spare cash turned negative.
It is a strange thing to find reassuring, but it explains a lot. When the richest company in the world blinks at the bill, you understand why the race is narrowing to a handful of players who can still afford to run it.
Plenty else happened. Anthropic launched Claude Opus 5 at half the price of its top model, Samsung signed a chip partnership with Broadcom worth more than 200 billion dollars through 2030, and someone fit a working AI onto an 8 dollar chip that writes stories with no internet at all. We cover one of these every morning in the daily brief, in plain language, in about a minute. Catch it tomorrow.