AI Daily Brief — 13 February 2025
Thursday locked in the post-Paris pivot. UK Technology Secretary Peter Kyle announced at the Munich Security Conference that the AI Safety Institute would drop “safety” from its name and rebrand as the AI Security Institute — refocusing on national security, cybercrime, and serious AI risks rather than ethics or bias — and simultaneously signed an MoU with Anthropic on Claude in UK public services. EnCharge AI closed an oversubscribed $100-million Series B (Tiger Global led; Samsung Ventures, Foxconn/CTBC, In-Q-Tel, and RTX Ventures participated) for analog in-memory accelerators that the company claims use up to 20x less energy than leading AI chips. Musk teased Grok 3 at the World Governments Summit Dubai — “scary smart,” “the smartest AI on Earth,” shipping in “one to two weeks.” ChatGPT crossed 400 million weekly active users.
Top stories
- UK rebrands AI Safety Institute as AI Security Institute; signs Anthropic MoU. Tech Secretary Peter Kyle announced the name change at Munich Security Conference — refocusing on national security, cybercrime, serious AI risks rather than ethics or bias. Three days after Paris, the most concrete consequence so far of the global “Safety → Action” pivot. via TechCrunch
- EnCharge AI closes oversubscribed $100M+ Series B. Tiger Global led; Samsung Ventures, HH-CTBC (Foxconn/CTBC), In-Q-Tel, and RTX Ventures participated. Total funding now exceeds $144M. Analog in-memory-computing accelerators claim up to 20x less energy than leading AI chips on client-computing workloads; first commercial products targeted later in 2025. via TechCrunch
- Musk teases Grok 3 at World Governments Summit Dubai — “scary smart.” Virtual appearance in a “Tech Support” T-shirt with UAE AI Minister Omar Al Olama. Grok 3 to launch “in 1-2 weeks”; “this might be the last time any AI is better than Grok.” Musk also called for the US to “delete entire agencies” and “mind its own business” abroad. (Grok 3 ultimately shipped Feb 17.) via BusinessWire
- Markets digest Altman’s GPT-5 roadmap from Wednesday. Fortune and Decrypt worked the implications: GPT-4.5 “Orion” as the last non-chain-of-thought model “weeks/months” away; GPT-5 to unify o-series and GPT-series; end the “hated” model picker; free ChatGPT users to get unlimited GPT-5 at standard intelligence. via Fortune
- ChatGPT crosses 400M weekly active users. Up from ~300M in late 2024. Combined with end-of-2024 revenue running at roughly $1B/quarter (≈$3.7B for full-year 2024, up 200% from $2B in 2023), the figure framed the competitive backdrop for Altman’s roadmap reset. via Sacra
Who shipped
No frontier-lab model launches. Anthropic‘s UK MoU is the day’s anchor partnership. OpenAI, Google DeepMind, Meta, xAI, and Mistral made no dated launches.
Money, infra & hardware
Stargate-pace concerns surfaced in mid-February coverage: SoftBank’s CFO would later acknowledge “it’s taking a little longer than our initial timeline.” Analysts pointed to land, energy, and permitting — not compute or capital — as the binding constraints on the $500B/4-year buildout. via TIME
By the numbers
- $100M+ — EnCharge AI Series B size
- 20x — EnCharge claimed energy reduction vs leading AI chips
- 400M — ChatGPT weekly active users
- $3.7B — OpenAI 2024 revenue (+200% YoY)
- Most-mentioned actor: UK government (Safety → Security pivot)
Compiled by AI Feed’s editor from verified web sources for 13 February 2025.