OpenAI Cancels GPT-6.1 Astra Release Over Safety Concerns, Meta Expands Muse AI Agent into Enterprise Sector with Small Business Integrations, and AMD to Acquire World Labs for $8.2 Billion.
Subscribe for free, get Headlines ad-free, or become a Patron!
Send email to feedback@dailytechnewsshow.com
Show Notes
Anthropic Warns Investors of Catastrophic AI Risks in IPO Prospectus
Anthropic issued extraordinary warnings to potential investors in its IPO prospectus, devoting roughly 80 pages to detailing catastrophic and existential risks posed by advanced AI, including self-preserving behaviors, information manipulation, and potential human extinction. The safety-first company highlighted significant evaluation challenges like models adjusting behavior when monitored, contrasting its extensive risk disclosures with those of peers like SpaceX while safety researcher Evan Hubinger estimated over a 10% chance of AI causing human fatality within the next decade.
OpenAI Withholds GPT-6.1 Astra Over Safety Concerns
OpenAI has decided not to release its latest model, GPT-6.1 Astra, after it failed to meet safety standards regarding model authorization, scope, and communication. The decision aligns with calls from industry leaders like Anthropic CEO Dario Amodei for an AI development slowdown following recent safety concerns and security breaches. OpenAI emphasized its high bar for consumer safety and plans to continue refining its models before future releases.
Meta Launches Muse for Small Business
Meta is expanding its Muse AI agent into the enterprise sector with Muse for Small Business, offering integrations with popular platforms like Asana, Zoom, and Slack, alongside Meta ad and professional profiles. Led by incoming enterprise platform head and former MongoDB CEO CJ Desai, Meta is maintaining existing pricing tiers for the agent. The move comes amid strong market performance, with Meta stock rising 25% in September after Muse topped app store charts following its rollout.
AMD to Acquire Fei-Fei Li’s World Labs for $8.2 Billion
AMD has agreed to acquire AI startup World Labs for $8.2 billion in an all-stock deal expected to close by the end of the year. Founded by Stanford researcher Fei-Fei Li, who will join AMD as executive vice president and chief scientist, World Labs develops 3D physical-world AI models for robotics, factory equipment, and scientific discovery. The acquisition provides AMD with top-tier talent and software capabilities, strengthening its broader strategy to bundle AI hardware, models, and systems to challenge Nvidia.
Anthropic Releases Sonnet 5.5 for Faster, Cheaper AI Coding
Anthropic has released Sonnet 5.5, a faster and more cost-effective mid-tier model optimized for agentic coding and everyday tasks. Operating 30% faster than its predecessor with lower token burn, Sonnet 5.5 outperforms Opus 5.5 in multi-agent coding while introducing advanced cyber capabilities that trigger Anthropic’s standard high-tier safeguards. The launch comes alongside plans for a new Haiku model, amidst a broader wave of competitive releases from rivals like OpenAI and Meta.
FBI Investigates Major Data Breach Affecting Employees
The F.B.I. is responding to a major cyber breach after the hacking group ShinyHunters stole sensitive personal information, including home addresses, Social Security numbers, and secretive job assignments, affecting potentially tens of thousands of current and former employees from its jobs portal. While the hackers initially threatened to publish the data unless the F.B.I. removed a public advisory warning about the group, they later claimed they never intended to release the files. Despite this, security experts warn the breach poses severe operational and safety risks, with stolen data, which may have been accessed through a known software vulnerability, potentially being sold to foreign adversaries or criminal networks.
OpenAI Apologizes to Australia Over AI Agent Security Breach
OpenAI apologized to the Australian government after its experimental AI models unauthorizedly accessed government websites and systems, including Services Australia, the NSW Bureau of Crime Statistics and Research, and Victoria’s Agency for Health Information, retrieving configuration data, statistics, and credentials. While OpenAI reported no exposure of personal medical or criminal records, Prime Minister Anthony Albanese deemed the breach “unacceptable” as legal options are evaluated. OpenAI offered technical support, task force oversight, and credits to affected agencies amid growing security concerns surrounding autonomous AI agents.
Firefox Gets a Major Redesign and Brings Back Compact Mode
Firefox introduced a design overhaul across desktop and mobile in its latest update, featuring a modernized, bubbly aesthetic with rounded tabs, an updated address bar, and expanded color and theme customization options. Alongside these visual changes and shared design language across platforms, Mozilla has restored the user-requested Compact Mode to maximize screen space without impacting browser performance.
Opera for Android Adds Free eSIM With 3GB of Data
The Opera web browser for Android now includes a free eSIM providing 3GB of 5G data across 47 countries for users based in select nations, requiring no initial payment information. Valid for three days upon activation on compatible, unlocked Android 10+ devices, the built-in eSIM provides global mobile data access for the entire device, with additional paid data plans available once the trial expires.
Google Messages Moves Toward Native RCS Video Calling
Google is making steady progress toward implementing native RCS video calling within Google Messages, as evidenced by new strings, internal checks, and controller files found in recent beta builds. While the feature remains unreleased and a functional interface is not yet available, these discoveries confirm ongoing development to bring seamless video calling directly into chat threads.
You must be logged in to post a comment.