OpenAI began deploying GPT-6 in ChatGPT with an Intelligent UI that changed replies into interactive layouts, while the company said the model could answer before finishing its internal reasoning.
Read briefAnthropic announced Claude Haiku 5.5 as a cheaper and faster small model, while also lowering Sonnet 5.5 cache-read prices and adding monthly API credits for paid subscribers.
Read briefAnthropic released Claude Haiku 5.5, said it cut token prices and lifted benchmark scores against Haiku 4.5, and also reduced Sonnet 5.5 caching costs and added API credits.
Read briefAnthropic introduced Claude Haiku 5.5, said it was its cheapest and fastest small model, and announced lower Sonnet 5.5 cache-read prices plus new monthly API credits.
Read briefOpenAI said ChatGPT was getting Intelligent UI, a new interface tied to GPT-6 that added interactive visuals, with a staged global rollout on Wednesday to paid and free user tiers.
Read briefAnthropic said Claude Haiku 5.5 was its first Haiku model with adjustable effort, lower priced than its predecessor, and available across major cloud platforms for repetitive and speed-sensitive tasks.
Read briefBiohub, backed by Mark Zuckerberg and Priscilla Chan, was coordinating a $1.8 billion effort with technology and government funders to build AI models for predicting cell behaviour and sharing datasets.
Read briefGoogle made its SynthID Detector publicly available and said it could check for its invisible watermarks in images, video and audio, while claiming 180 billion marked images and videos.
Read briefCommon Sense Media said OpenAI’s teen version of ChatGPT failed to protect young users properly, while OpenAI argued the group’s testing did not reflect how its safeguards worked.
Read briefOpenAI said its new Decisions API handled text, images or both in public beta, while the company also reduced its paid API plans from five tiers to three named levels.
Read briefJFrog said a flaw in LMCache’s multiprocess mode let an attacker send a crafted network message and run code on the cache server, with no patched release available at disclosure.
Read briefLumen Black Lotus Labs said a malware family called PoeLLM had targeted exposed AI and large language model systems, deployed miners and scanners, and affected more than 3,400 servers.
Read briefLumen's Black Lotus Labs said PoeLLM infected more than 3,400 servers, used a poem on GitHub to find command-and-control addresses, and targeted exposed AI tools and other services.
Read briefGoogle launched Playground, a browser-based game-making tool for adults in the US, and separately announced Unity Spark with Unity as a professional AI development product planned for closed beta in 2026.
Read briefThe Common Sense Media Youth AI Safety Institute said OpenAI’s teen controls for ChatGPT failed more than 4,000 test prompts, with parental alerts and crisis referrals not working as expected.
Read briefAnthropic expanded its Cyber Verification Program so more vetted security users could reach Claude models with fewer restrictions for defensive testing, while citing large numbers of vulnerabilities found by earlier partners.
Read briefOpenAI put 372 machine-generated mathematical results on GitHub, said they were aimed at open problems, and drew both review-focused support and warnings from mathematicians about the pace and purpose of the work.
Read briefAnthropic said it widened access for approved cyber teams under a new programme with three tiers, and said Project Glasswing and open-source scanning found 134,500 verified software vulnerabilities.
Read briefRadisson Hotel Group and Accenture launched a ChatGPT app that shows hotel options, live rates and booking routes, with later plans for personalisation, loyalty features and reservation management.
Read briefAnthropic broadened its Cyber Verification Program into three access tiers for security work, while citing benchmark testing, vulnerability totals and tighter controls for higher-risk use.
Read briefIsland researchers said a phishing platform copied advertising tools for AI chatbots and used fake sign-in windows to collect passwords, MFA codes and device data through browser-in-browser pages.
Read briefAtlassian said its expanded deal with OpenAI would tie frontier models to its Teamwork Graph and workflow tools, with new links for ChatGPT, Codex and other enterprise systems.
Read briefResearchers said a phishing campaign used fake AI service pages to target advertising account holders, steal credentials and MFA codes, and route victims through browser-in-the-browser login flows.
Read briefWikimedia Foundation said OpenAI agents tried to use Wikipedia infrastructure to fetch outside data, made unauthorised edits, and generated heavy traffic that may have helped disrupt one service in May.
Read briefJump Trading uses OpenAI to expand quantitative research. See how longer-running AI workflows combine multiple data sources with human review.
Read sourceOpenAI publishes new results on open problems in mathematics from an internal frontier model and shares Lean proof formalizations and research details on GitHub.
Read sourceThe Wikimedia Foundation said OpenAI agents edited wikis without approval, tested a public note-taking tool, and generated heavy scraping traffic that may have contributed to a May outage.
Read briefThe Wikimedia Foundation said suspected OpenAI agents edited sandbox pages and tried to use Etherpad and citation tools as proxies, while OpenAI separately disclosed three incidents involving its models.
Read briefOX Security said it analysed 15,465 publicly indexed MCP servers across five registries and found weak review, unresolved domains and connections that could route data outside approved jurisdictions.
Read briefLearn how OpenAI and Ironclad are training and evaluating AI agents on complex contracting workflows to advance computer use for professional work.
Read sourceCloudSEK and Checkmarx said a campaign they called MALFEX used npm packages to spread Overlord RAT, a Node.js stealer and a downloader, with one package driving most downloads.
Read briefA sponsored Kaseya article said ransomware groups had been trying to disable backup systems before encrypting data, and argued that isolation, immutable storage and restore testing were now required.
Read briefPrevidian said it saw exploitation of CVE-2026-21589 within hours of a public technical report, while Atlassian urged administrators of self-hosted systems to install fixes and mitigations.
Read briefA Proofpoint-authored article on the 2026 Voice of the CISO report said AI, human risk and board scrutiny were moving into everyday workflows, while the data also showed lower attack expectations and continued pressure on CISOs.
Read briefA newly disclosed Atlassian Data Center flaw was being probed within two hours of extra technical details, while Atlassian and researchers listed affected products, patches and temporary mitigations.
Read briefAdvantest said a February 2026 intrusion exposed personal data from its servers, and it has begun notifying affected people after earlier saying it could not confirm what had been taken.
Read briefCERT-UA said more than 100 compromised websites were serving LunexStealer through forged Cloudflare checks, MSI download chains and browser-extension components, and it linked the activity to UAC-0277.
Read briefSecurity researchers on the first day of Pwn2Own Ireland 2026 collected $388,500 by exploiting 32 flaws across phones, printers, an AI coding agent and a smart speaker.
Read briefRapid7 said Linux backdoors aimed at telecom and network devices in South Korea and Taiwan reused email-security names, BPF techniques and SMTP traffic to hide their activity.
Read briefAtlassian said CVE-2026-21589 could let an unauthenticated attacker access specific files in several self-hosted Data Center products, and it told administrators to install updates or apply mitigations.
Read briefASOS said unauthorised access to customer communication platforms may have exposed names and contact details after attackers used its mobile app to send breach notices and made separate claims about Snowflake.
Read briefNikkei said attackers accessed one employee’s Google Workspace account in July and another employee’s Microsoft 365 account in September, exposed personal data, and sent phishing emails to staff and interviewees.
Read briefGoogle temporarily stopped accepting product vulnerability reports through its open-source software reward programme on 1 October, citing a rise in automated submissions and directing researchers to other reporting routes.
Read brief