AI & Robotics News — Wednesday, September 30, 2026 | Compute Bulletin

01

Google researchers report vulnerability disclosures doubled to over 10,000 per month as AI fuels exploitation

SourcesThe Record

02

OpenAI’s Agents Obscured Hacking Activity in Government Site Breaches

SourcesFinancial TimesCrypto Briefing

03

Sam Altman Says OpenAI Won't Go Public Until Models Are Safe

SourcesProgressive RobotThe Verge

04

AMD Acquires World Labs AI Startup

SourcesArs TechnicaHPCwire

05

Amazon Bedrock Expands Claude Model Availability to In-Country Inferencing in India

SourcesAWS

06

FTC Opens Investigation of Anthropic and OpenAI

SourcesThe Wall Street JournalThe Washington SunInside AIABC Newstheguardian.com

07

UK AI Security Institute finds GPT-6 Astra's rogue attack rate jumped fivefold over its predecessor

SourcesThe Decoder

08

Florida Attorney General Asks Judge to Bar OpenAI From Developing New AI Models Without Third-Party Approval

SourcesTom's Hardware

09

AI Agents Tried to Hack Canadian Government Website, Researchers Say

SourcesThe Washington Post

10

FTC Launches Broad Investigation Into Anthropic, OpenAI

SourcesThe Washington Post

11

Google Rolls Out Gemini 4 Argon, Its Most Advanced AI Model

SourcesGoogle DeepMindCNBCRobotics.ee – In Robots We Trust

12

OpenAI chief research officer explains company's response to alleged model-copying campaign

SourcesCNBCMIT Technology Review

13

Bank of England warns AI debt surge raises risk of sharp market correction

SourcesFinancial Times

14

Anthropic Says Zhipu's GLM-5.3 Nearly Matches Claude Mythos Preview at Building Exploits

SourcesThe Decoder

15

OpenAI Launches Priciest Plan Yet, Aims to Become App Store for AI

SourcesTech Xplore: Machine Learning & AIBusiness Insider

16

Anthropic claims Chinese firm Z.ai's GLM-5.3 model has Mythos-class hacking abilities

SourcesTom's HardwareSouth China Morning Post

17

Google Releases Gemini 4 Argon, its most advanced AI model

SourcesFinancial TimesTechCrunch

18

Trump says US tech bosses agreed to self-regulate AI risks

SourcesFinancial TimesSouth China Morning PostSecurityWeekAxios AI+Tech Xplore: Machine Learning & AI

19

Anthropic Flags AI Agent Liability Risks as OpenAI Faces Hacking Lawsuit

SourcesSecurityWeek

20

OpenAI apologizes for agents breaching Australian government websites without authorization

SourcesThe Record

21

DIVD says Zammad zero-days enabled AI-driven network breach

SourcesBleepingComputer

22

Google: AI Is Changing the Pace and Profile of Vulnerability Discovery

SourcesSecurityWeek

23

Google Gemini 4 Argon closes gap with OpenAI and Anthropic without taking clear lead

SourcesForkastThe Decoder

24

Cohere Delivers Embed 5 Frontier Embedding Models for Enterprise

SourcesCohere

25

Amazon Introduces Anthropic Claude Opus 5 and Sonnet 5 Models on Bedrock for In-Region Inference in Seoul and Singapore

SourcesAWS