AI Industry News Brief — October 19, 2026

Collection window: UTC 2026-10-17 20:47 to 2026-10-18 20:47 (last 24 hours; KST 2026-10-18 05:47 to 10-19 05:47). Publish times were checked on each article page and converted to UTC. Stories already covered in the previous brief (Replit–Anthropic, the Nova Scotia data center, DeepSeek V4.1-Flash-1017) were excluded. The Korean publishers’ ‘No Books, No AI’ appeal (Oct 17, 15:37 UTC) and the Claude Code 2.1.302 release (Oct 16) fell before the cutoff, and The Verge’s AI-agent blocking story, the Telegraph’s Microsoft–Nscale report and NBC’s piece on Claude’s values were left out because their exact publish times could not be verified. A quieter weekend: 8 stories.

Anthropic commits 5% of its shares to a new Public Benefit Fund, with 2% more for every 50% rise in value

Anthropic announced it will commit part of its equity to the newly created Anthropic Public Benefit Fund. Subject to stockholder approval, 5% of shares are committed now, and the board may add 2% for every 50% increase in company value above a set baseline. The Fund is a wholly owned LLC, and none of its Anthropic equity may be used for lobbying or political giving, in the US or abroad. Combined with the co-founders’ pledge of 80% of their equity and an employee equity-match program (about half of staff participate), roughly 20% of Anthropic’s shares will be dedicated to public benefit.

Reference: Anthropic (2026-10-18 16:49 UTC)

Disney hires most of bankrupt Character.AI’s workforce, about $60M in annual pay

After Character.AI filed for bankruptcy earlier this week, Disney agreed to make job offers to at least 60 employees, the majority of its workforce, with estimated annual compensation of around $60 million. A legal filing reported by Bloomberg says the vast majority of those offered jobs accepted. Disney had earlier agreed to buy some Character.AI assets for $20 million, but that is a stalking-horse bid, so another company could still top it.

Reference: TechCrunch (2026-10-18 09:30 PDT = 16:30 UTC)

Sam Altman: OpenAI will build a humanoid robot, targeting a 2027 demo

OpenAI CEO Sam Altman said on the Sources podcast that the company will build a humanoid robot. He argued the real challenge is the AI ‘brain’ that controls the robot, not the body. A demo is expected in 2027, but design, pricing and release timing have not been disclosed. OpenAI is hiring for robotics roles in actuators, motion control and manufacturing, while Tesla and XPENG are already further along with their own humanoids.

Reference: Engadget (2026-10-17 23:45 UTC)

Schwab and JPMorgan block outside AI agents from moving customer money

Charles Schwab CEO Rick Wurster told Bloomberg TV that Schwab blocks unauthorized AI agents from executing transactions for clients, seeing tens of such attempts every week. Schwab steers customers to its own assistant, Charley, and gives outside LLMs read-only MCP server access that excludes moving money. JPMorgan CEO Jamie Dimon reportedly said his firm also blocks outside agents from customer account data. With no federal statute governing agent liability in financial services, big firms are building a de facto framework through exclusion.

Reference: Forkast (2026-10-17 21:11 UTC)

Three unannounced ‘aurora’ models appear in OpenAI’s API, with no official word

OpenAI’s API model list now shows three new identifiers: aurora-red-alpha, aurora-green-alpha and aurora-blue-alpha. OpenAI has not announced or documented them, and has said nothing about their capabilities, pricing or intended use. The attention traces back to a single post on X by AI researcher @scaling01, which also referenced an earlier ‘aurora-alpha’ model that first surfaced on OpenRouter in February 2026. Any link to a GPT-5.3-Codex-Spark variant remains unconfirmed speculation.

Reference: Crypto Briefing (2026-10-18 03:18 UTC)

Meta drops terms language that would have let it train AI on business customers’ websites

Meta recently revised its commercial terms to add wording allowing it to use content from business account holders’ websites to develop and train products, including AI models. Legal experts quoted by Nieman Lab warned this could enable scraping of news publishers’ sites. A Meta spokesperson said the update didn’t change how the company crawls the web, and The Verge later reported the provision had been removed. Meta says the crawling provisions apply only to its advertising and business offerings and has clarified that in the terms.

Reference: TechCrunch (2026-10-17 14:31 PDT = 21:31 UTC)

Kimi K3 contributor Chen Guangyu, 17, leaves Moonshot AI for the US

Chen Guangyu, born in 2009, announced on X that he is leaving Chinese AI startup Moonshot AI for the San Francisco Bay Area. He joined the Kimi research team at 16 and worked on Kimi Delta Attention and attention-residual techniques that were central to the Kimi K3 model unveiled in July. He said he would not join Anthropic despite an offer from one of its researchers, prompting speculation he may enroll at a US university. The move highlights intensifying competition for AI talent in China.

Reference: Seoul Economic Daily (2026-10-18 18:02 KST = 09:02 UTC)

Samsung taps TSMC’s 3nm process for Nvidia’s custom HBM base dies

Samsung Electronics is working with TSMC on the base die for Nvidia’s custom NVHBM memory. Nvidia reportedly prefers TSMC’s 3nm process, with yields above 90%, versus roughly 60% for Samsung’s 2nm SF2 process. UBS projects 2027 HBM demand to rise 91% from this year, leaving a supply shortfall of about 10%, and Samsung’s own base-die capacity is limited. Samsung is also preparing to make base dies on its own 2nm process.

Reference: Seoul Economic Daily (2026-10-18 17:59 KST = 08:59 UTC)

Today’s Summary

Frontier labs are widening their governance and business footprint (Anthropic’s Public Benefit Fund, OpenAI’s humanoid robot, unannounced models surfacing in the API).
Boundaries around AI agents and data are being drawn (Schwab and JPMorgan blocking outside agents, Meta withdrawing its training clause).
The race for talent and chips continues (Disney absorbing Character.AI staff, a 17-year-old Kimi researcher heading to the US, Samsung–TSMC HBM cooperation).