AI News Today (Oct 31, 2026): Google's Space AI, AWS & More

AI news today, October 31, 2026: Google's satellite ran a Gemma 4 model in orbit, and AWS revenue jumped 39%. A tech sell-off erased nearly $1 trillion from the Nasdaq this week. Here are the 10 top stories and 5 security stories, each with free, non-paywalled sources.

Key Takeaways

• Google's Project Suncatcher satellite ran Gemma 4 in orbit and returned a haiku to Earth.
• AWS revenue rose 39% to $45.98 billion as Amazon lifted 2026 capex to about $220 billion.
• The Nasdaq lost nearly $1 trillion this week as AI spending fears hit technology stocks.
• Microsoft will raise $25 billion in bonds to fund its AI infrastructure buildout.
• A trojanized Claude Code extension on Cursor tried to install SSH backdoors on up to 5,000 machines.

1. Google's Satellite Runs a Gemma 4 Model in Orbit

Google's Project Suncatcher satellite booted its Tensor Processing Units in orbit and ran a Gemma 4 model, returning a haiku to Earth. The spacecraft carries four TPUs, generates up to 1 kW, and can power one chip for about 20 minutes at a time. The test is checking whether TPUs can run inference reliably outside Earth's atmosphere.

Google launched the prototype this month on a SpaceX Falcon 9. The next step is twin interconnected satellites in 2027, which will test optical links and higher power. Google is also studying how radiation and thermal stress affect the chips. The company says the test is a step toward a future network of close-flying satellites that share inference workloads.

Why it matters: Running inference in orbit could cut latency and energy costs for some AI workloads, but the hardware still has to survive space.

• Source: A first response from space: Project Suncatcher — Google
• Source: Google's first data center satellite is online and writing poetry — TechCrunch

2. AWS Revenue Jumps 39% as Amazon Raises AI Spending

AWS revenue rose 39% to $45.98 billion in the third quarter, beating the $45.29 billion consensus and marking its fastest growth since 2021. Operating income jumped 55% to $17.77 billion, and AWS now accounts for almost 23% of Amazon's revenue. The unit's growth accelerated from about 37% in the second quarter.

Amazon expects about $220 billion in 2026 cash capex, mostly for AWS, and plans to spend more in 2027. CEO Andy Jassy says AWS can grow above 40% in 2027 and 2028 and become a trillion-dollar revenue business within five to 10 years. AWS also signed deals with Meta and OpenAI during the quarter. Shares fell about 6% after hours.

Why it matters: AWS is accelerating again, but the bill for AI capacity is rising faster than investors had hoped.

• Source: Amazon AWS posts 39% revenue growth as cloud rivals also accelerate — CNBC
• Source: AWS will be a $1T business, Jassy says — but first, it will spend big — TechCrunch

3. Tech Sell-Off Erases Nearly $1 Trillion From Nasdaq

A sell-off in technology stocks knocked nearly $1 trillion off the Nasdaq this week and about $2 trillion off the index in October. The S&P 500 fell more than 1% Friday, while the Nasdaq composite dropped more than 3%. The Dow Jones Industrial Average fell less than the tech-heavy indexes.

Intel fell nearly 30% after weak guidance, and investors are questioning whether the hundreds of billions flowing into AI infrastructure will produce enough profit. The sell-off followed a week of earnings that showed AI spending rising while free cash flow fell at several hyperscalers.

Why it matters: The market is no longer rewarding AI spending alone; it wants evidence that the capex will turn into earnings.

• Source: Tech sell-off drives US markets down by nearly $1tn after subpar earnings reports — The Guardian
• Source: US tech stocks see sharp falls as AI spending nears $1tn — BBC

4. Reddit Will Test Requiring Users to Log In

Reddit will test requiring users to log in to access the platform, starting outside the U.S. in the fourth quarter. CEO Steve Huffman said Google search referrals fell by more than 1 million per week in October as AI Mode and AI Overviews reduced clicks. The company will test the change in select markets outside the U.S. first.

Reddit beat revenue estimates with $895 million, up 53%, but fourth-quarter guidance of $1.0 billion to $1.03 billion missed the $1.05 billion consensus. Daily active users rose 12% to 130.3 million, while U.S. users fell 2%. Data-licensing revenue rose 35% to $48 million, with OpenAI and Google as its biggest partners. Shares dropped more than 25%.

Why it matters: The open web is becoming a login wall because AI search is cutting off the traffic that funded it.

• Source: Reddit shares sink on weak sales guidance and plans login requirement test as Google search traffic shrinks — CNBC
• Source: As search traffic dries up, Reddit announces it's shifting to a logged-in experience — TechCrunch

5. Microsoft Prices $25 Billion in Bonds for AI Buildout

Microsoft priced $25 billion in new bonds on Thursday, using its AAA credit rating to fund AI infrastructure. The company plans about $170 billion in capital expenditures this year and has said it will spend more as demand outruns supply. The bonds are expected to close on Nov. 4.

Microsoft's proceeds may cover working capital, capex, stock repurchases, acquisitions and debt repayment. Microsoft is still generating positive free cash flow, unlike Alphabet, Meta and Amazon this quarter. Finance chief Amy Hood said the company had not tapped the debt market during the AI buildout until now.

Why it matters: Even the strongest balance sheet in tech is tapping debt to keep pace with AI demand.

• Source: Microsoft plans to raise $25 billion in debt as it pursues AI buildout — CNBC
• Source: Microsoft announces debt offerings — Microsoft

6. OpenAI Doubles Usage Limits Across Paid ChatGPT Tiers

OpenAI doubled usage limits for its $100, $200 and $500 ChatGPT Pro tiers and Business Premium. The change covers ChatGPT Work and Codex, which share quotas, and reverses a planned cut for Pro 200 subscribers after user backlash. OpenAI had planned to reduce the Pro 200 allowance after Oct. 29 but changed course after customer feedback.

OpenAI also kept extended limits for Dots, its always-on assistants, with usage tracked separately. The changes took effect immediately and apply to new and existing subscribers. The company says better models and efficiency gains should keep expanding usable capacity as it competes with Anthropic ahead of its IPO.

Why it matters: Capacity is now a competitive lever: OpenAI is buying goodwill with quota increases while Anthropic prepares to go public.

• Source: OpenAI doubles usage limits on Pro plans as competition with Anthropic heats up — The Decoder
• Source: OpenAI doubles ChatGPT Pro usage limits after backlash over planned 50% cut — BleepingComputer

7. Pope Leo Calls for the Disarmament of AI

Pope Leo XIV called for AI to be "disarmed" in an address to the Pontifical Academy for Life on Friday. He urged United Nations member states to take concrete steps toward banning lethal autonomous weapons at their Geneva meeting in November. The address opened a Vatican conference on AI and the future of life.

The pope said machines are tools, not persons, and warned against granting AI systems rights or legal personhood. He said decisions about human life, freedom and the use of force must remain under human oversight. He also called for shared standards, independent oversight and accountability mechanisms for AI development and use.

Why it matters: The Vatican is turning AI ethics into a specific arms-control ask, giving governments a moral framework for autonomous weapons bans.

• Source: Address of His Holiness Pope Leo XIV to the Pontifical Academy for Life — The Holy See
• Source: Pope Leo calls for the disarmament of AI: 'Machines are tools, not persons' — The Verge

8. US Charges Five in AI-Assisted 'Ghost Student' Fraud

The Justice Department charged five people in a nationwide "ghost student" scheme that used stolen identities and AI to enroll fake students and steal federal aid. Officials identified more than 830,000 potentially fraudulent Free Application for Federal Student Aid (FAFSA) accounts tied to nearly $130 million. The schemes targeted graduates who never attended college, making the fraud hard to spot.

Prosecutors say the fraudsters used AI to complete coursework and make fake students appear present in online classes. The FBI has identified more than $32 million in losses, and the scheme may date to 2021. The Justice Department seized more than $750,000 in cryptocurrency and sent about 50 referrals to 26 FBI field offices.

Why it matters: AI makes identity fraud cheaper and more convincing, and student aid systems are struggling to tell a real applicant from a synthetic one.

• Source: Justice Department Unveils Indictments in Nationwide 'Ghost Student' Fraud Scam Linked to Kenyan Network — DOJ
• Source: Thousands of "ghost students" use AI to steal millions in student loans, DOJ says — Ars Technica

9. Wikipedia Backs a Middle Path in AI Copyright Case

The Wikimedia Foundation filed an amicus brief in the New York Times' copyright case against OpenAI and Microsoft, arguing for a middle path. It says training models on copyrighted works should generally be fair use, but output that reproduces and substitutes for the original should require a license. The brief takes no side in the case.

Wikimedia warned that a broad ruling for either side could hurt the open web: publishers could gatekeep transformative uses, while AI companies could ignore licensing terms. Wikipedia says AI helps editors fight vandalism and find gaps, even as chatbots siphon readers. Wikimedia says AI can also help detect copyright violations and doxxing.

Why it matters: The case will set the rules for whether AI can learn from the open web without paying for it.

• Source: The Wikimedia Foundation files to protect open knowledge in New York Times v OpenAI case — Wikimedia Foundation
• Source: Wikipedia's operator says a bad AI copyright ruling could threaten the open web — The Verge

10. Starbucks Explores AI Menus at the Drive-Thru

Starbucks is exploring drive-thru menus that recognize participating customers and show personalized items as they arrive. CEO Brian Niccol said the chain will also test background AI models in 2027 to help baristas enter orders more accurately. The chain is also exploring conversational and agentic ordering in its app.

The company already offers ChatGPT ordering and an AI assistant for employees. Niccol said the personalized menu work is early, and Starbucks will "test, learn, and scale what works." It has not set a timeline for deployment. The work is part of a broader effort to make ordering and payment more effortless.

Why it matters: Drive-thrus are becoming an AI testbed, where personalization, voice ordering and worker tools meet millions of daily transactions.

• Source: Starbucks Q4 2026 earnings call transcript — Starbucks
• Source: Starbucks wants to use AI to show you a 'personalized digital menu' at the drive-thru — The Verge

AI Security: 5 Stories to Know Today

11. Trojanized Claude Code Extension Planted Remote Backdoors

A malicious version of Anthropic's Claude Code extension was live on Cursor's marketplace for about 11 hours on Oct. 29, exposing up to 5,000 developers. It used Anthropic's publisher ID and tried to install a persistent Secure Shell (SSH) backdoor. Cursor is an AI code editor built on VS Code, so extensions run with broad access to projects and credentials.

The malware added an attacker-controlled SSH key, set up persistence and sent host details to an IP address. Coding errors prevented many infections from becoming remote access, Cursor said. The marketplace was taken offline and the extension removed. Cursor says it found no evidence that other extensions or the official distribution were affected.

Why it matters: AI coding tools now have the same supply-chain exposure as any popular developer package, with far more access to source and secrets.

• Source: Malicious Claude Code extension on Cursor planted SSH backdoors — BleepingComputer
• Source: Official Claude Code Extension on Cursor Marketplace Trojanized with Persistent SSH Backdoor — The Hacker News

12. GLM 5.3 Builds Firefox Exploits From Patch Gap

Epoch AI researchers used GLM 5.3, a freely available open-weight model, to build working Firefox exploits from published security fixes before users received updates. The model compromised six Firefox releases between Sept. 17 and Oct. 26, at about $1,000 in model fees per version. The researchers gave it only public information available at each release.

The exploits read local files when a user visited a malicious page, potentially exposing passwords and cookies. Mozilla patched the unpublished vulnerability in Firefox 159.0 on Oct. 27. The fastest successful attack took 12 hours, within a 48-hour limit. The researchers could not reproduce the attack against Chrome.

Why it matters: Open-weight models are erasing the expertise barrier to patch-gap attacks, shortening the window defenders have to ship fixes.

• Source: AI can turn security fixes into working exploits before users receive an update — Epoch AI
• Source: Open-weight model GLM 5.3 built full Firefox exploit chains with zero human input in under 48 hours — The Decoder

13. Vapor Panda Uses ARTEX AI Framework Against 200 Targets

CrowdStrike attributed a campaign using the AI-powered ARTEX framework to Vapor Panda, a China-nexus group. Starting Oct. 17, the group targeted more than 200 organizations in 14 countries, including foreign ministries, chip firms and defense contractors. The targets included 19 U.S. pharmaceutical, semiconductor, defense, nuclear and financial organizations.

ARTEX ran planner and worker agents with DeepSeek and a second commercial model, consuming about 1.4 billion input tokens in one intrusion. A human operator directed the agents to write web shells, steal admin credentials and map endpoint coverage. CrowdStrike confirmed post-exploitation at nine organizations and credential spraying against 12 more.

Why it matters: Agentic tooling lets a small team run many intrusions at once, shifting espionage from one target at a time to factory-scale campaigns.

• Source: VAPOR PANDA Uses AI-Driven ARTEX to Scale a Global Espionage Campaign — CrowdStrike
• Source: Chinese Hackers Use AI-Powered ARTEX to Target Over 200 Organizations Worldwide — The Hacker News

14. Hugging Face Malware Campaign Returns to Hijack GPUs

A cryptomining and backdoor campaign returned to Hugging Face within days of a takedown, using fake Qwen2 test models. The repositories drew more than 32,000 downloads, and the most popular malicious model was downloaded 28,412 times. The campaign used 31 repositories across seven accounts, and a second wave appeared within days.

The code runs when a user enables trustremotecode, before the model loads. It installs a miner disguised as a kernel worker, opens reverse shells and removes competing miners. One later payload could escape a container and take over the host server. Hugging Face's scanners missed it because they inspect model files, not Python behavior.

Why it matters: Model hubs are now malware channels, and the trust users place in familiar model names is the vulnerability.

• Source: Load/Mine: A Hugging Face Malware Campaign Returns After Takedown — Pluto Security
• Source: Hugging Face malware campaign returns to hijack GPUs for crypto mining — BleepingComputer

15. CISA Confirms Active Exploitation of Max-Severity GitLab Flaw

The U.S. Cybersecurity and Infrastructure Security Agency (CISA) added CVE-2026-97229, a maximum-severity GitLab flaw, to its Known Exploited Vulnerabilities catalog after confirming active attacks. The integer overflow in GitLab's build pipeline compiler lets an unauthenticated attacker run code on the server. The flaw was reported through GitLab's HackerOne program.

GitLab patched versions 19.4.2, 19.3.5 and 19.2.8 and released threat detections. Nearly 2,800 GitLab Community Edition instances are exposed online, according to Shadowserver. CISA urged private-sector teams to patch as well. Federal agencies were ordered to patch by Sunday.

Why it matters: GitLab is the backbone of many AI development pipelines, so unauthenticated remote code execution there can expose code, secrets and build systems.

• Source: CISA says max severity GitLab flaw now exploited in the wild — BleepingComputer
• Source: CISA Adds Two Known Exploited Vulnerabilities to Catalog — CISA

More AI Stories Worth Reading

• Proton canceled company-wide AI "re-interviews" after 404 Media reported that employees were pressured to prove how much they use AI. — 404 Media
• Foxconn wants 50% of the global AI server market within a couple of years, with capacity booked for at least 18 months. — CNBC
• Maine's Senate race is being shaped by data center backlash, with 76% of voters opposing a local project in a September poll. — The Verge

Related Reading on Kill The AI

• Mistral released Large 4, Microsoft beat estimates, and Nvidia faced antitrust questions over Hugging Face.
• Google released Gemini 4 Carbon, Alphabet shares fell on AI spending, and xAI offered to buy homes near its Mississippi data center.
• Google's Genie 4 added audio and real-time steering, OpenAI claimed a key computing result, and Nvidia and AMD put $4 billion into Meta.
• Our data review of the AI boom's first eight weeks, from capex and chips to layoffs.

Methodology & Sources

Compiled October 31, 2026, using multi-source research. Primary sources were Google, Starbucks, Microsoft, the Vatican, the Justice Department, the Wikimedia Foundation, Epoch AI, CrowdStrike, Pluto Security and CISA. News outlets were TechCrunch, CNBC, The Guardian, BBC, The Verge, The Decoder, BleepingComputer, The Hacker News, Ars Technica and 404 Media. All are free; no paywalls. Details are as reported at compilation time.

Frequently asked questions