AI Weekly
AI News ·
This week in AI — Gemini 4 Argon, Anthropic heads to Singapore, scam detection that stays on your phone
Official announcements only, Oct 12–17, 2026. Google's new top model and its launch pricing, Anthropic's fifth Asia-Pacific office, and an on-device model that warns you mid-scam.
We only cover what companies published on their own blogs and newsrooms, and every item links to the original. Benchmark claims below are the companies' own numbers, not independent tests.
1. Google launches Gemini 4 Argon (Oct 13)
Google released Gemini 4 Argon, now its most capable model, aimed at long, multi-step work like coding, finance, law and science.
- Who gets it: Google AI Pro and Ultra subscribers in the Gemini app and AI Mode in Search. Developers via the Gemini API in Google AI Studio.
- Context: up to 1M tokens of combined input and output through the API.
- Launch pricing: $2 per 1M input tokens and $10 per 1M output tokens until November 30, 2026, then $4 and $20.
- The claims: #1 on the Vals Index (finance, coding, legal, tax) and a 0.7% attack success rate on an indirect prompt-injection benchmark — both reported by Google.
What it means for you: If you already pay for Gemini, you can switch to it today. If you build on the API, the launch window is the cheap time to test it.
Source: Gemini 4 Argon — Google
2. Anthropic opens in Singapore and signs a government MOU (Oct 14)
Anthropic, the company behind Claude, opened its fifth Asia-Pacific office in Singapore and signed a memorandum of understanding with the Ministry of Digital Development and Information.
- The deal covers: AI safety, security and model evaluation; research on AI adoption; and responsible use across the public and private sectors.
- Usage: In Anthropic's Economic Index, Singapore ranks 2nd of 121 countries for Claude.ai use relative to population.
- In government: more than 1,500 engineers and technical staff at GovTech Singapore have adopted Claude Code.
What it means for you: No new features. It's another sign that AI labs are moving fast into government partnerships across Asia.
Source: Anthropic opens Singapore office — Anthropic
3. A scam detector that never leaves your phone (Oct 14)
Google DeepMind and GovTech Singapore built a lightweight Gemma model fine-tuned on local scam conversations — fake banks, fake officials, fake public figures.
- A small classifier screens messages first. Suspicious conversations go to the fine-tuned model, which looks at the whole exchange and shows a real-time warning before you send money or personal details.
- It runs on the device, only with the user's consent. The announcement says neither the Singapore government nor Google can read the conversations.
- The rollout will be phased. No date yet.
What it means for you: Expect more "AI that stays on your phone" features — useful exactly where you'd least want your messages sent to a server.
Source: On-device AI scam detection — Google DeepMind
The takeaway
Models keep getting stronger, AI companies are moving into government work, and more AI runs privately on your own device. The model matters less than how you use it — find out your AI type in 3 minutes.
Sources
- Gemini 4 Argon: Built to solve your hardest problems — Google (2026-10-13)
- Anthropic opens Singapore office and signs MOU with the Government of Singapore — Anthropic (2026-10-14)
- Google DeepMind and Singapore Government collaborate to build fine-tuned AI model — Google DeepMind (2026-10-14)
Written with AI (Claude) and checked against the linked sources.