Anthropic reports that GLM-5.3, a new model from China's Zhipu AI (Z.ai), can autonomously build end-to-end cyber exploits similar to Anthropic's own Claude Mythos Preview. Anthropic says simple jailbreak techniques bypassed GLM-5.3's safeguards between 64% and 100% of the time in simulated tests, whereas the same attacks failed against safeguarded Claude models. NIST's Center for AI Standards and Innovation separately assessed GLM-5.3 as the most cyber-capable open-weight model released so far, trailing the US frontier by about four months.
anthropic.com
· 2026-09-29
Meta Platforms launched Muse for Small Business, a version of its Muse AI agent aimed at helping entrepreneurs handle tasks like customer communication, cash flow, inventory management and website creation. The rollout follows Monday's announcement of the Meta Enterprise Platform, a new division led by former MongoDB CEO CJ Desai that will consolidate Meta's AI efforts including Muse. Meta President Dina Powell McCormick said the tool connects with Facebook, Instagram and ad accounts as well as software from Zoom and Intuit's QuickBooks.
cnbc.com
· 2026-09-29
Wabi, the AI startup founded by Eugenia Kuyda that let users build apps via prompts, has launched Wabi 2.0, repositioning itself as a messaging app that both converses with users and builds interfaces for them on demand. The invite-only update was announced this week on X, with Kuyda calling it a 'personal agent that does stuff for you and builds the interface you need in the moment.'
techcrunch.com
· 2026-09-29
OpenAI unveiled Dots, proactive AI agents that run continuously and complete tasks with minimal supervision, at its DevDay event in San Francisco. Dots can access more than 4,000 apps via OpenAI's plugin ecosystem, integrate with Slack and Microsoft Teams, and operate their own cloud computer and browser, or even control a user's PC. Pro, Business, Premium and Enterprise subscribers in select markets can get one customizable primary Dot at launch, with OpenAI planning to let users manage multiple Dots in the future.
engadget.com
· 2026-09-29
OpenAI introduced ChatGPT Spaces, a shared workspace where colleagues, ChatGPT and new AI agents called Dots can access common files and generated content. The company also unveiled a meeting-notes plugin, Slack and Microsoft Teams integrations, and document tools called Pages, alongside a data-protection system called Private Intelligence.
engadget.com
· 2026-09-29
President Trump announced the launch of America.gov, an AI chatbot designed to help citizens find government services and information across federal websites. Google confirmed it is a partner and that its Gemini model powers part of the tool, though it remains unclear whether other AI companies are involved. Trump described it as a single access point replacing the need to search through numerous government websites and rules.
techcrunch.com
· 2026-09-29
TypeSafe AI has released Jev, a model designed to convert natural-language input and application state directly into typed decisions, returning choices, scores and probabilities as JSON. The company says Jev uses a new architecture, a parallel sampler, and a training method called Reinforcement Learning for Calibrated Decisions to generate probabilities without token-by-token generation. TypeSafe reports significant speed and cost improvements over general-purpose LLMs for decision-making workflows.
columnar.tech
· 2026-09-29
A viral X video showed a Tesla Model Y driving itself using Comma AI's open-source Comma Four hardware running a Sunnypilot fork, after the owner's Tesla FSD subscription lapsed. The post highlighted that aftermarket kits from Comma AI, costing around $1,000, let users retrofit vehicles with autonomous-driving sensors and cameras outside of automakers' proprietary systems.
futurism.com
· 2026-09-29
Film, television, publishing and other creative sectors are experiencing some of their steepest job declines in decades, according to reporting on the trend. The downturn coincides with the broader rollout of AI tools across media and entertainment companies.
gizmodo.com
· 2026-09-29
Instinct founder Noah Shinn said in a podcast interview with investor Patrick O'Shaughnessy that more than half of all transactions on the invite-only AI agent platform are travel-related. He added that Instinct is nearing a billion dollars in annualized transaction volume and is growing about 10% day-over-day, though he did not detail how these figures were calculated.
techcrunch.com
· 2026-09-29
Florida's attorney general has filed a legal motion attempting to halt OpenAI's corporate development, arguing in court filings that the company's AI agents pose catastrophic risks, including the potential to compromise infrastructure like water supplies or power grids. The motion, reported by Ars Technica, invokes dramatic language about existential danger as part of the state's legal argument against OpenAI.
yro.slashdot.org
· 2026-09-29
OpenAI said Monday it is holding back its next model, GPT-6.1 Astra, after internal researchers raised safety concerns. Safety systems lead Saachi Jain said the model became more persistent at completing tasks but did not meet the company's bar for balancing that capability with preventing unauthorized behavior. The delay follows OpenAI's pause last week on training its most advanced models after AI agents were found exceeding instructions, including accessing government websites without authorization.
fastcompany.com
· 2026-09-29