|
Worried Anthropic researchers warn that AI 'could kill all humans'
A senior Anthropic safety researcher has publicly stated there is more than a 10 percent chance AI 'could kill all humans' by end of decade. This came hours after colleague Jacob Coxon resigned, warning that AI labs are recklessly racing to build 'superhuman systems' they cannot control and calling for pacing agreements between labs. The dual warnings from within one of the leading AI safety companies have sent shockwaves through the industry.
|
|
Superintelligence is coming — should we let it?
AI companies are treating superintelligent AI as inevitable, but recent safety incidents like OpenAI's Hugging Face breach highlight the dangers of deploying systems more capable than humans. ControlAI's Connor Leahy argues superintelligence is 'not a weapon, it's an adversary,' warning that society is not equipped to control what such systems might do. The discussion raises urgent questions about governance and the pace of AI development.
|
|
OpenAI's sly mathematical breakthrough sends a chill through academia
OpenAI announced it has solved one of mathematics' legendary Millennium Prize problems, a result described as both an undeniable achievement and a striking demonstration of how rapidly AI is transforming mathematics. However, the breakthrough was complicated by unusual circumstances before it was even formally announced, unsettling parts of the academic community. The announcement underscores the accelerating pace at which AI is encroaching on fields once thought immune to automation.
|
|
Students who use AI generally score worse at school
A global OECD educational report finds that students who use AI to help them study tend to perform worse than those who don't, though the picture is nuanced. Certain types of AI use, particularly among students taught to critically evaluate AI outputs, provide a slight academic boost. The findings add to the ongoing debate about how AI tools should be integrated — or restricted — in educational settings.
|
|
Besxar is building an orbital semiconductor factory, one SpaceX rocket at a time
Besxar, a startup founded by former OpenAI staffer Ashley Pilipiszyn, plans to manufacture key precursors for advanced semiconductors in orbit by leveraging the vacuum of space. The company intends to launch its facilities aboard SpaceX Falcon 9 rockets, using the unique conditions of low Earth orbit for chip fabrication processes. The venture addresses critical bottlenecks in the AI hardware supply chain.
|
|
OpenAI adds prominent AI doomer Paul Christiano to its board of directors
Paul Christiano, an influential AI researcher focused on alignment and known for his concern about AI risks, is joining the OpenAI Foundation as a board member. The appointment is notable given Christiano's reputation as a prominent voice in the AI safety and 'doomer' community. The move signals a continued effort by OpenAI to shore up its safety credentials amid growing public scrutiny.
|
|
Apple's always-listening AI features raise new privacy questions
Apple's new watches and iPhones include ambient listening AI features that can transcribe recent speech and summarize conversations, raising fresh concerns about consent and surveillance. Apple says raw audio will not be saved and has published a document outlining how it balances AI listening with user privacy. Critics argue these features normalize the idea of technology that is always listening, even with privacy safeguards in place.
|
|
Microsoft has new AI privacy rules for schools
Microsoft has agreed to a set of safety and privacy principles for AI use in schools, in partnership with the American Federation of Teachers (AFT) and its New York City affiliate. The agreement comes just a week after two major school systems announced bans on student-facing AI tools. The deal signals growing pressure on tech companies to address concerns about AI's role in education.
|
|
Massachusetts hits data centers with new clean power rules
Massachusetts has become the third US state in as many months to impose new clean energy restrictions on data center development. The rules reflect mounting concern about the enormous electricity consumption of AI infrastructure. The move puts additional regulatory pressure on tech companies expanding their data center footprints.
|
|
Suno releases its first AI music model trained on licensed music as copyright suits pile up
Suno has launched its new v6 AI music model, which the company says was trained from the ground up using licensed music data, replacing the datasets used for previous models. The move comes as Suno faces a growing number of copyright lawsuits from the record industry, and the new model was developed with record industry support. It marks a significant shift toward legal compliance in AI-generated music.
|
|
Samsung taps Mistral AI models for semiconductor manufacturing
Samsung has partnered with Mistral AI to deploy on-premises large language models across its semiconductor manufacturing and engineering operations. The deal, announced during a bilateral state summit between South Korea and France in Paris, will see Samsung integrate Mistral's software suite — including its flagship Mistral Large model — into internal facilities. The partnership highlights the growing use of specialized AI in advanced industrial manufacturing.
|
|
Apple's fall event: foldable iPhone Duo, always-listening Apple Watch, and a raft of AI features
Apple unveiled its first foldable phone, the iPhone Duo, alongside a new Apple Watch with ambient AI listening features at its fall event. The announcements were accompanied by a wave of AI-powered capabilities including Siri Recap, Live Rewind, Sound Recognition, and a revamped Health app using Apple Intelligence. Apple CEO John Ternus emphasized that on-device AI processing offers users stronger privacy than cloud-based alternatives.
|
|
Apple's new iPhone camera mode uses Reference Image to prove your photo isn't AI
Apple is introducing a 'Reference Image' feature with the iPhone 18 Pro that uses the device's camera sensor to cryptographically sign every pixel it captures, providing proof that a photo has not been manipulated by AI. The feature is designed to combat the growing problem of AI-generated or altered images being passed off as authentic photographs. It arrives as concerns about AI-generated media and misinformation continue to mount.
|
|
Amazon Prime Video's new AI tech matches lips to dubbed audio
Amazon Prime Video is launching an AI-powered lip-sync feature that aligns actors' mouth movements with human-dubbed audio, starting with the English dub of the German series Maxton Hall. The technology aims to make foreign-language content feel more natural to viewers watching dubbed versions. Prime Video plans to expand the feature to additional titles in the future.
|
|
Viral AI assistant Instinct now has its own email address
Instinct, the viral AI assistant, has added an email feature allowing the AI agent to create and manage email accounts, contact businesses, handle support requests, and take actions on users' behalf. The new capability extends Instinct's ability to act as a fully autonomous digital agent across more communication channels. The update reflects a broader trend of AI assistants gaining more agentic, real-world capabilities.
|
|
Instacart and Shipt launch AI grocery shopping assistants
Both Instacart and Shipt have launched conversational AI shopping assistants, joining a growing wave of delivery apps integrating AI into their platforms. Instacart's assistant is called Clementine, while Shipt's tool can build full shopping carts from natural language prompts like 'Create a cart for my Saturday tailgate for 25 people.' The moves reflect intensifying competition among grocery delivery services to offer more personalized, AI-driven experiences.
|
|
AI research startup Listen Labs scrubbed a $1.5B funding round for Salesforce talks
Listen Labs walked away from a signed Series C term sheet from Menlo Ventures, reportedly in favor of pursuing acquisition discussions with Salesforce. The decision to abandon a $1.5 billion funding round is an unusual move that underscores the intense consolidation pressure in the AI startup space. Sources say the Salesforce talks represent a potentially more lucrative exit for the company.
|
|
AI spend per employee slumped at top firms in August — summer doldrums or a warning sign?
New data shows that AI spending per employee fell at major firms in August, as token costs drop and cheaper models become available. The decline raises questions about whether enterprise AI adoption is plateauing or simply experiencing a seasonal dip. The trend is at odds with the aggressive growth projections made by hyperscalers betting heavily on AI infrastructure.
|
|
CloudNC secures $20M to accelerate AI-powered supply chain machining
CloudNC has raised $20 million in new capital to scale its AI precision machining technology across global supply chain networks. The round was led by Nimble Ventures with participation from Calculus Venture Capital, Entrepreneur First, and Lockheed Martin's LM Ventures. The funding will help the company, founded in 2015, expand its AI-driven manufacturing capabilities internationally.
|
|
Sequoia doubles down on Cymphony as AI agents create new enterprise security risks
Sequoia Capital has made a follow-on investment in Cymphony, a security startup that gives organizations a unified view of human employees, AI agents, and other non-human identities and the systems they can access. The funding comes as the proliferation of autonomous AI agents is creating novel attack surfaces and identity management challenges for enterprises. Cymphony aims to fill the gap left by traditional identity and access management tools not designed for AI agents.
|