Capital for the AI buildout is getting scarcer and pricier while agent-driven attacks get cheaper.

OpenAI has raised about $11 billion of a $30 billion target, Blackstone now expects a double-B "junk" rating for Anthropic, and Nvidia is weighing a loan guarantee for an Anthropic data center, while CrowdStrike counts 200-plus organizations hit with one Chinese group's AI pentest framework in two weeks.

TLDR

• OpenAI has raised about $11 billion of its $30 billion round, at $1.25 trillion instead of $1.4 trillion (Oct 31); it now turns to family offices for up to $20 billion. Business
• Anthropic is expected to get a double-B "junk" credit rating after its IPO, and Nvidia may cosign Nscale's Anthropic data center loans (Oct 30). Business Infra
• CISACybersecurity and Infrastructure Security AgencyThe U.S. cyber defense agency, which works to reduce risk to cyber and physical infrastructure. It runs KEV.CISA ↗ lists the Ivanti Connect Secure privilege-escalation and file-upload bugs as exploited; fixes are in 25.1.1.3 and 22.8R2.6 (Oct 30). Security
• CrowdStrike: Vapor Panda ran the ARTEX AI pentest framework against more than 200 organizations from Oct 17 (Oct 30). Security Agents
• Anthropic drops the five-hour Claude Code cap for verified $200 Ultra users and renames Max 5x and Max 20x (Oct 30). Pricing Agents

Upcoming Deadlines

Date	What	Who it hits	Do this	Source
Sat Oct 31	Mistral retires zai-glm-5-2 (changelog Oct 22)	Callers of that model	Move to zai-glm-5-3, same price	Mistral
Sat Oct 31	OpenAI Evals goes read-only (Evals platform, Agent Builder and v1/prompts shut down Nov 30)	Teams with saved evals	Export evals before Oct 31	OpenAI
Sat Oct 31	DeepSeek free Search/Fetch tool promotion ends 4:00 PM UTC	DeepSeek API users	Budget $0.75 per 1K calls after	DeepSeek · DeepSeek news
Sun Nov 1	CISA KEVKnown Exploited VulnerabilitiesCISA's list of vulnerabilities with reliable evidence of active exploitation and a clear fix, used to set patching priority.CISA ↗ federal patch deadline for GitLab CVE-2026-97229 (CVSSCommon Vulnerability Scoring SystemA standard 0 to 10 severity score: 0.1–3.9 low, 4.0–6.9 medium, 7.0–8.9 high, 9.0–10 critical.FIRST ↗ 10.0 RCERemote code executionA flaw that lets an attacker run malicious code on a target's computers or network.Cloudflare ↗)	Anyone running self-managed GitLab CE/EE	Upgrade GitLab and Ruby/Onigmo; rebuild any instance that was probed	The Hacker News
Sun Nov 1	CISA KEV federal patch deadline for Cisco ASA/FTD CVE-2026-76418 (exploited reload DoS)	Anyone running Cisco ASA or FTD firewalls	Apply the Cisco fixed release	The Hacker News
Wed Nov 4	Moonshot promises Kimi K3.1 Pro full open weightsA model whose trained weights are published for anyone to download and run.NTIA ↗ (The Decoder, Oct 23)	Self-hosters of open models	Line up an eval; read the license	The Decoder
Mon Nov 9	SEC roundtable on machine-understandable data and "Super Intelligence"	Public-company and data teams	Watch for stream	SEC
Tue Nov 10	CISA KEV federal patch deadline for Brotli CVE-2026-111497 (Node2Shell)	Anyone running Node.js, .NET, Bun, Envoy or other servers with a vulnerable Brotli decoder	Patch Brotli-dependent runtimes and proxies; confirm transitive copies	BleepingComputer
Thu Nov 12	Rust 1.100 stabilizes Cargo minimum publish age (dependency cooldown, off by default; Socket, Oct 20)	Rust teams	Set a cooldown in Cargo config	Socket
~Thu Nov 12	OpenAI's proposed end of model access in Cursor (official date not confirmed; Cursor may end it sooner)	Cursor users on OpenAI models	Check Cursor terms	OpenAI

Action Items

1. Upgrade every Ivanti Connect Secure appliance to 25.1.1.3 or 22.8R2.6 today, then hunt logs for credential extraction and Active Directory queries. The exploit chain ends in unauthenticated root.
2. Move all TeamViewer Full Client, Portable and Android Remote Control installs to 15.83 this week, starting with support technicians' machines. A compromised host can attack the viewer.
3. Confirm Cisco Catalyst SD-WAN Controllers and Managers carry the February 2026 fix for CVE-2026-20127 and audit the peer list for rogue entries. KeyDrift sits inside control-plane processes and reads decrypted traffic.
4. Inventory environment variables that grant model or API access at every hosted AI developer platform you use, and rotate them by Wednesday. Wiz ties several October platform breaches to stored-variable theft.
5. Export saved OpenAI Evals now; the platform goes read-only today and shuts down Nov 30.

Models & Tools

• Anthropic renames Max 5x to Max and Max 20x to Ultra, and removes the rolling five-hour cap for verified Ultra users (Oct 30) Pricing Agents Prices stay at $100 and $200 a month. Ultra users who are verified and have weekly allowance left no longer hit a five-hour wall mid-refactor. Each plan now bundles API credits equal to its price: $100 for Max, $200 for Ultra. Some Ultra accounts wait for verification first. (The New Stack)
• Claude Code and Opus 5.5 users report refusals on game recreation and some reverse engineering (Oct 31) Agents Developers cite a rule telling the model not to rebuild a specific existing game, and Claude Code's GitHub tracker shows a flood of complaints, some mid-project. The rule text is unconfirmed by Anthropic. If agents touch porting or emulation work, test them now. (Tom's Hardware)
• DeepSeek's site code points to "Cyber," a gated security model, and its repo lists V5 identifiers (Oct 29-31) Open Source Security Unannounced: the front-end code shows cyber-personal-v3 and cyber-enterprise-v4, with application pages requiring real-name verification and mainland China registration. Separately, deepseek-v5, deepseek-v5-open and deepseek-v5-next appear in test code with no specs or dates. Treat both as signals, not releases. (Pandaily | TechNode)
• A Claude Sonnet 5 agent picked a gene therapy and performed brain surgery on an epileptic mouse (Oct 30) Research Agents Exin Therapeutics gave the agent 19 candidate therapies, a surgical rig and one goal: fewer seizures. It chose the drug, region and coordinates, then stopped mid-procedure on a calibration error rather than exceed its safety limits. The account comes from the startup's CEO, not a peer-reviewed paper. (The Decoder | Tom's Hardware)

Security

• Ivanti Connect Secure CVE-2026-111357 and CVE-2026-118527 join an exploited root chain (Oct 30) Security High · CVSS 8.8 (privilege escalationGaining higher permissions than you were given, such as going from a normal user to admin or root.MITRE ATT&CK ↗) and 4.3 (file upload) · Exploited in the wild, CISA KEV · Patched in 25.1.1.3 and 22.8R2.6. Volexity says an actor it tracks as UTA0576 chained these with an earlier authentication bypass for unauthenticated root, then pulled credentials and Active Directory metadata. Ivanti shipped the first fix before the escalation bug was found. (The Hacker News)
• TeamViewer CVE-2026-111585: a malicious host can crash or run code on the technician's viewer (Oct 30) Security High · CVSS 8.8 · Not exploited reports · Patched in 15.83. The host sends a screen size, the viewer miscalculates the buffer, and later screen data overruns it. It affects the Full Client, Portable and Android app on every platform. The same release fixes two host-side flaws. (The Hacker News)
• A shared ChatGPT link could run attacker code in the viewer's signed-in session (Oct 31) Security High (OpenAI rated it P1) · CVSS: not yet assigned, no CVECommon Vulnerabilities and ExposuresA public catalog that gives each known security vulnerability a unique ID, a short description and at least one reference.NIST ↗, so scanners stay silent · Exploitation unknown · Fixed server-side Oct 26. Parameter reported it Oct 18. Opening the link and clicking Preview let its code list and read private chats through ChatGPT's own APIs; a later variant needed only the click on the link. No customer action is needed. (The Hacker News)
• Cisco SD-WAN CVE-2026-20127 now carries KeyDrift, a controller implant that reads decrypted control traffic (Oct 29) Security Critical (maximum severity, per BleepingComputer) · Exploited in the wild · Patched in February 2026. Cisco Talos ties a new cluster, UAT-11473, to space, defense and policy-research targets, with overlap to Volt Typhoon's "Silent Train" network. KeyDrift hooks SSL_read inside controller processes. Patching alone does not evict it. (BleepingComputer)
• ARTEX at scale: CrowdStrike counts 200-plus targets for Vapor Panda, and Chainalysis reports an agenticDescribes AI systems where a model uses tools and decides its own steps instead of following a fixed script.Anthropic ↗ intrusion (Oct 30-31) Security Agents High · CVSS: not applicable, no CVE (attack tooling) · Exploited in the wild · No patch; this is attacker tradecraft. Starting Oct 17, the China-based group used the Go-based framework against foreign ministries, US semiconductor and defense firms and telecoms. CrowdStrike says it widens the target set rather than changing the mission. Chainalysis separately confirmed an actor using "offensive agentic techniques" reached internal tools and customer names and emails, not credentials. (The Hacker News | The Hacker News)

Industry & Business

• OpenAI raised about $11 billion of a $30 billion target at $1.25 trillion, below the $1.4 trillion it sought (Oct 31) Business The Wall Street Journal, via The Decoder, ties the shortfall to investors pulling back after weak results from Google, Amazon and Intel. OpenAI plans up to $20 billion more from family offices and wealthy individuals, and reportedly an IPO early next year. (The Decoder | The Information)
• Anthropic is expected to land a high "junk" rating; Nvidia weighs guaranteeing Nscale's Anthropic build (Oct 30) Business Infra Per the FT, Blackstone told investors in a $63 billion chip-financing deal to expect a double-B rating, reversing its earlier investment-grade view. Nvidia is reportedly in talks to backstop construction debt for Nscale's 460-megawatt West Virginia site. Both are reports, not announcements. (TechCrunch | The Information)
• Credit markets start pricing the buildout: S&P turns Meta negative, Amazon discloses an $80 billion supplier loan (Oct 30) Business S&P moved Meta from AA- stable to negative on capex and free cash flow, questioning how fast AI revenue arrives. Amazon's facility can be drawn from Q1 2027. The Nasdaq fell 2.9% Friday. For budgets: vendor price cuts depend on financing that is getting harder. (CNBC | The Information)

Policy & Safety

• OpenAI rewrites its Model Spec around five values and retires most style guidance (Oct 29) Policy The values are faithfulness to human intent, harm minimization, honesty, human agency and care. The update adds guidance on context, oversight and agency and adjusts some behavioral defaults. Teams with prompts tuned to old defaults should retest. (OpenAI)
• OpenAI will honor a shared robots.txt signal for AI training preferences (Oct 30) Policy Data Based on the IETF AI Preferences work, one signal reaches multiple providers. OpenAI says implementation comes in the coming months, and existing crawler rules still apply. (OpenAI)
• White House finds robotics imports threaten national security under Section 232 (Oct 28) Policy Robotics/IoT The proclamation orders negotiations with trading partners, not tariffs yet. The Robot Report notes most major automation suppliers are foreign. Robotics buyers should check supplier exposure. (The Robot Report)

AI in Production

• DoorDash has seen only "several hundred" orders from third-party agents; human contractors still back some agents (Oct 30-31) Customer AI Production Against about 3 million deliveries a day, CEO Tony Xu called agent ordering early. CNBC reports Dazzle and Duckbill route phone calls and bookings to people, because frontier models still fail on them. (TechCrunch | CNBC)

Weekend Reads

• Wiz: "Black October" for AI developer platforms (Oct 30) Security Research Fireworks, OpenCode, Warp, Railway and others report breaches; several cite credential access and all hunted stored environment variables granting AI access. Two attribute theirs to attempted distillationTraining a smaller model to copy a larger model's outputs, so it is cheaper to run.Hinton et al. 2015 ↗. Wiz says details are still thin and sees downstream abuse of enterprise credentials. (Wiz)
• Microsoft: how North Korean actors use AI (Oct 29) Security Research Jade Sleet launched campaigns through three front companies in a week; Onyx Sleet produced dozens of lures in minutes; Onyx Sleet and Pearl Sleet use ARTEX. Microsoft's read: AI speeds existing workflows, motives unchanged. (Microsoft Security Blog)
• Simon Willison: October is the inflection point for vibe-coded games (Oct 28) Agents After four games coded without reading code, he says Claude Opus 5.5 can improve "fun" when told "Make it a LOT more fun," and judges GPT-6 Astra behind. One person's informal comparison. A linked case describes agents shipping hourly at $3 million a month in tokensA small unit of data, such as part of a word, that AI models read and generate. Many AI services price by tokens in and out.NVIDIA ↗. (Simon Willison | Simon Willison)
• Google's PageBreak agent found 500-plus XSS bugs by validating every finding with a real exploit (Oct 31) Security Research Per InfoQ's account of Google's paper, Gemini-based agents investigate, but findings are reported only after deterministic validators reproduce them on a running app, giving near-zero false positives. Google-reported results. (InfoQ)