Collection window: UTC 2026-10-21 20:48 to 2026-10-22 20:48 (last 24 hours; KST 2026-10-22 05:48 to 10-23 05:48). Publish times were checked on each article page and converted to UTC. Stories already covered yesterday (Claude Duet, Flapping Airplanes, etc.) are excluded. Walmart’s MCP agent gateway (19:30 UTC), Recursive Flint (20:27 UTC) and Google’s Gemini 4 Argon usage figures (15:13 UTC) fell before the cutoff; Mistral Large 4 (official page dated Oct 6) and Nvidia’s supply-chain expansion (no time shown) were left out because their publish times could not be confirmed.
DeepSeek reportedly raises about $15B, two days after US federal agencies were told to remove it
According to Bloomberg, China’s DeepSeek has raised about $15 billion (roughly 100 billion yuan), one of the largest rounds ever in China, with CATL and Tencent named among potential investors. The report came two days after the US ordered federal civilian agencies to strip DeepSeek technology from their systems. DeepSeek raised more than $7 billion at a $50 billion valuation in June and is reportedly weighing a China IPO as early as next year. The company has not confirmed the round.
Reference: TechCrunch (2026-10-22 06:11 PDT = 13:11 UTC)
China-linked hackers used AI agents to breach ConnectWise and hit 2,250 organizations at ~3,000 sign-ins per minute
Microsoft attributed the breach of ConnectWise’s cloud backup service to Storm-0940, a China state-aligned group. The attackers stole application credentials linking ConnectWise to customers’ Microsoft 365 environments, letting them log in as authorized users and read email and files. Reconnaissance, password spraying and OAuth app abuse were automated with ARTEX, a Go-based agentic AI penetration-testing framework, hammering about 2,250 tenants at nearly 3,000 sign-in attempts per minute. Microsoft said that speed left defenders “little time to detect and respond”; ConnectWise paused the backup service and is restoring it in phases.
Reference: TechCrunch (2026-10-22 09:47 PDT = 16:47 UTC) · The Hacker News (2026-10-22)
OpenAI proposes a five-stage disclosure framework for AI-discovered vulnerabilities
Drawing on its response to the PHP2Shell flaws (CVE-2026-109618 and -109619), OpenAI proposed a coordinated process for disclosing bugs found by AI.
- Maintainers and researchers confirm the bug and build and test a fix, including bypass checks.
- Affected organizations get early notice to prepare backports, mitigations and WAF rules.
- Government agencies are brought in.
- The circle widens to more distributors, providers and operators.
- After the patch ships, guidance is published and operators are supported.
The timeline speeds up if a flaw is exploited or leaks, and maintainers make the final call on early access. OpenAI expects AI-driven findings to “increase sharply in the coming months” and is asking for community input on disclosure-window length and related questions.
Reference: Help Net Security (2026-10-22 07:35 UTC)
Meta posts $64B Q3 revenue and creates a non-voting share class to fund its AI buildout
Meta’s Q3 revenue rose 26% year over year to $64 billion, but costs and expenses jumped 68%. CFO Susan Li confirmed a new class of non-voting stock, to be distributed to existing shareholders as a stock dividend and used for capital raising, employee pay and acquisitions while preserving Class A and B voting power. Zuckerberg said more than 10 million people in the US and Canada set up Muse personal agents in about six weeks, and highlighted Muse Spark Open (open weights) and the Muse Code coding agent.
Reference: TechCrunch (2026-10-21 15:45 PDT = 22:45 UTC)
Uber rolls out ChatGPT Work to all 30,000 employees
Uber expanded its OpenAI partnership to give its roughly 30,000 employees worldwide access to ChatGPT Work, OpenAI’s enterprise agent. An earlier pilot with about 1,000 staff across finance, operations and product found that around 80% of early adopters were non-engineers, using it mainly to analyze operational data, localize training materials and build personalized dashboards. Uber did not disclose the cost and plans to track usage and outcomes to guide further AI adoption.
Reference: TechCrunch (2026-10-22 09:00 PDT = 16:00 UTC)
Ex-OpenAI science lead Kevin Weil’s Warp Sciences raises $215M to automate drug-optimization work long outsourced to China
Warp Sciences, founded by Kevin Weil, who led OpenAI for Science, raised $215 million led by Lightspeed, with IVP, Premji Invest and Lux Capital joining, at an $865 million post-money valuation. It is building a fully automated US lab for the step where chemists make hundreds or thousands of variants of a promising molecule. Drugmakers have long outsourced that work to Chinese firms such as WuXi AppTec, a dependence that US-China tensions have made riskier. Its chemists will wear cameras and pressure-sensing gloves to generate training data for Warp’s AI models.
Reference: The Next Web (2026-10-22 16:05 UTC)
1Password tops $500M ARR as usage by AI agents grows 500%
1Password’s annual recurring revenue passed $500 million, up from $400 million a year ago. Monthly active usage of its developer and AI security tools rose 200% among humans and 500% among autonomous AI agents. It added about 20,000 business customers to pass 200,000, and launched a new platform the same day to help customers deploy personal AI agents. As agents increasingly handle passwords and API keys, credential management for agents is becoming a growth engine.
Reference: BetaKit (2026-10-22 13:00 UTC)
Today’s Summary
- Chinese AI showed its financial muscle despite US sanctions, with DeepSeek’s reported $15B raise, while on the offensive side Storm-0940 used AI agents to attack at 3,000 sign-ins per minute.
- Defenders are adapting to AI speed too, from OpenAI’s five-stage disclosure framework to surging demand for agent credential management at 1Password.
- Meta’s new share class, Uber’s company-wide ChatGPT rollout and the Warp Sciences round show AI investment reshaping capital structures and industrial supply chains.
Leave a Reply