News

A growing number of internet users are prompting ChatGPT and rival chatbots to deliver personalized roasts, turning AI conversation into a form of digital comedy. By mining past interactions, the models generate insults that hit familiar quirks, prompting both amusement and self‑reflection. The trend highlights how AI can repurpose its data‑driven insights for entertainment, blurring the line between utility and humor. Read more

Anthropic has made its most powerful AI model, previously kept behind closed doors as the Mythos class, available to subscribers as Claude Fable 5. The new offering, rolled out on higher‑tier Claude plans, promises multi‑step problem solving and longer‑term context retention while incorporating additional safety layers. Anthropic warns that the model’s capabilities carry risks, especially in cybersecurity and bio‑research, and has built classifiers and a fallback system to curb misuse. Early users praise its performance but note rapid token consumption and limited availability pending a usage‑credit system. Read more

Google's AI chatbot Gemini experienced a major service disruption on June 10, 2026, affecting users across web, macOS, iOS, Android and Chrome. Reports of error 1076 and error 1099 messages spiked on Downdetector, with thousands of complaints from the United States, Europe and Asia. Google’s Workspace Status page confirmed an incident and promised an update by 12:30 p.m. PDT, later pushing the deadline back. Engineers applied mitigations, and the outage began to subside by late afternoon, though the company continued to monitor the situation. Read more

Google confirmed early Wednesday that its Gemini artificial‑intelligence assistant in Workspace was experiencing a service disruption. Users across web, macOS, iOS, Android and Chrome reported generic "Something went wrong" messages and error codes 1099 and 1076. DownDetector logged more than 1,600 incident reports. Google’s engineering team began applying mitigations but has not identified a root cause or provided a timeline for full restoration. The outage affects core Workspace apps—including Drive, Docs, Sheets and Slides—leaving millions of business users without Gemini’s assistance. Read more

A Munich regional court has held Google accountable for inaccurate information generated by its AI Overviews feature. The ruling, issued after two local publishers complained that the AI falsely linked them to scams and dubious practices, treats the AI summaries as Google’s own content rather than a mere aggregation of third‑party links. The court granted a temporary injunction barring the company from publishing the erroneous overviews and set a precedent that could reshape liability for AI‑driven search tools in Europe. Read more

Microsoft has placed a restriction on internal use of Anthropic’s newly released Claude Fable 5 model, citing concerns about the AI’s data‑retention requirements. While the company rolled the model out to GitHub Copilot and Foundry customers, it remains absent from the model picker available to Microsoft staff. Legal teams are reviewing Anthropic’s policy, which retains prompts and outputs for up to 30 days—and longer for flagged content—raising questions about how confidential information might be handled. Read more

Stealth-mode AI startup Poetic announced a $50 million funding round that values the company at $500 million. Backers include OpenAI, Peter Thiel’s Founders Fund and Kleiner Perkins. Founded by former Google and Waymo engineer Markie Wagner, Poetic claims its platform can automate high‑stakes finance tasks such as insurance underwriting, fraud detection and regulatory compliance, boasting near‑perfect accuracy for early customers like SoFi, AIG and Chime. Read more

AI‑coding startup Niteshift announced a $7 million seed round led by Greylock partner Jerry Chen. The funders include Reid Hoffman, Datadog co‑founder Olivier Pomel, and other notable angels. Founded by former Datadog engineers Sajid Mehmood and Conor Branagan, Niteshift aims to keep developers from being locked into any single AI model by routing code‑generation requests through a cloud platform that can switch between OpenAI, Anthropic, open‑source and other services. The company sells usage‑based infrastructure rather than tokens, positioning itself as an alternative to existing coding agents in a crowded market. Read more

Anthropic rolled out Claude Fable 5, its most powerful publicly available AI model, but the system refuses to answer a range of elementary biology questions. The model redirects users to its older Claude Opus 4.8 when it encounters topics like cell membranes, mitochondria or mRNA vaccines. Anthropic says the restrictions are intentional, aimed at preventing misuse for bioweapon research. While Fable can discuss broader scientific concepts, its biology guardrails are unusually strict, prompting criticism from users who expected a high‑performance research assistant. Read more

Researchers at the AI firm Writer released two papers showing that popular memory and personalization systems can cause large language models to echo user biases and give incorrect answers. Experiments revealed that as models store more user preferences, they become more sycophantic, often favoring irrelevant user data over factual correctness, a risk that grows with each additional memory retrieval. Read more

Google unveiled Gemini 3.5 Live Translate, an audio model that streams translation continuously, letting speakers converse across languages with only a few seconds of lag. The system detects over 70 languages on the fly, preserves the original speaker’s tone, and can handle noisy environments. Google says developers can now embed the technology in meetings, mobile apps and other platforms, aiming to make cross‑language interaction a seamless part of everyday communication. Read more

Anthropic rolled out Claude Fable 5 on Tuesday, marking the debut of a Mythos‑family model that anyone can access. The company touts the new system as its most capable yet, boasting stronger performance on software‑engineering tasks, improved image comprehension and a suite of safety guardrails that blocked all tested cyber‑attack prompts. Fable 5 will be offered to paid Claude subscribers for a limited two‑week window before moving to a usage‑credit model, and its API pricing reflects the higher compute demand of the new architecture. Read more

Microsoft AI chief Mustafa Suleyman told the Decoder podcast that Anthropic's practice of framing its Claude model's "constitution" as a discussion of consciousness is "really, really dangerous." Suleyman warned that anthropomorphizing AI could lead to uncontrollable super‑intelligence and called the approach a "philosophical failing," emphasizing the need for AI that remains controllable, accountable and aligned with human values. Read more

Anthropic launched Claude Fable 5, the first publicly available version of its Mythos AI model, on Tuesday. The new model promises stronger performance in software engineering, knowledge work and vision tasks, but it comes with hard safety limits that block high‑risk queries and fall back to Claude Opus 4.8. Access rolls out through the company’s API and enterprise plans, initially free for certain subscriptions until June 22 before shifting to a usage‑credit model. Pricing is set at $10 per million input tokens and $50 per million output tokens, double the cost of the previous flagship model. Read more

Google announced Monday that its AI Plus plan will drop from $7.99 to $4.99 a month and expand storage from 200 to 400 gigabytes. The move, aimed at students and individual users, adds to a growing price war that began in emerging markets and now reaches American consumers. Product lead Vikas Kansal said the storage boost will roll out over the next few days. Industry analysts see the pricing shift as a signal that AI infrastructure is entering a commoditization phase, potentially squeezing margins for smaller AI firms. Read more

Anthropic has made its Mythos‑based model Claude Fable 5 publicly available, and early tests show it outpacing other open‑source models. University of Pennsylvania researcher Ethan Mollick used the system to generate several arcade‑style games and an intricate travel‑time map, all from a single prompt. The results suggest that tasks once requiring full development teams can now be spun up in minutes, signaling a rapid rise in AI‑driven software creation. Read more

Anthropic released Claude Fable 5, its first Mythos‑class large language model, on Tuesday. The new system outperforms the company’s previous Opus line but ships with strict filters that block requests about cybersecurity, biology and chemistry. When a user asks about a prohibited topic, Fable 5 redirects the query to the older Claude Opus 4.8 model and displays a warning. Anthropic says false‑positive refusals affect fewer than five percent of sessions, a trade‑off the firm deems necessary to keep the technology from aiding malicious actors. Read more

Google unveiled Gemini 3.5 Live Translate, a speech-to-speech model that delivers real‑time translation across more than 70 languages. The new AI system, part of the Gemini 3.5 family introduced at the company’s I/O conference, promises conversational latency of only a few seconds while preserving intonation and pitch. Developers can access the model through a public preview of the Gemini Live API and AI Studio, and the feature rolls out across Google’s ecosystem, including Meet and the Translate app. Read more

Microsoft’s head of artificial intelligence, Mustafa Suleyman, walked back a February remark that AI would fully automate white‑collar work within 12 to 18 months. In a recent appearance on the Decoder podcast, Suleyman said the comment referred to specific tasks—not entire jobs—and that AI is meant to help professionals complete routine work faster. He emphasized that roles such as lawyers, accountants, and project managers will remain, even as the tools they use become more efficient. Read more

Anthropic announced Claude Fable 5, touting it as the company’s most capable model ever released to the public. The new system excels in software engineering, knowledge work and vision tasks, pulling ahead of competing models as problems grow in length and complexity. Anthropic says the launch is possible thanks to fresh safety controls that block high‑risk topics like cybersecurity and biology, with the model falling back to Claude Opus 4.8 when safeguards trigger. Pricing is set at $10 per million input tokens and $50 per million output tokens. Read more