Actualités — 2026-06-10

Affichage 6 articles Toutes les actualités

Google launches Gemini 3.5 Live Translate for real‑time multilingual conversations

Google launches Gemini 3.5 Live Translate for real‑time multilingual conversations CNET
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. Lire la suite

Anthropic launches Claude Fable 5, the first widely available Mythos model with enhanced safety

Anthropic launches Claude Fable 5, the first widely available Mythos model with enhanced safety CNET
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. Lire la suite

Microsoft AI chief warns Anthropic against treating Claude as conscious

Microsoft AI chief warns Anthropic against treating Claude as conscious The Verge
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. Lire la suite

Anthropic Opens Claude Fable 5 to the Public with New Safety Guardrails

Anthropic Opens Claude Fable 5 to the Public with New Safety Guardrails TechCrunch
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. Lire la suite

Google Cuts AI Plus Subscription to $4.99, Doubles Storage in U.S. Rollout

Google Cuts AI Plus Subscription to $4.99, Doubles Storage in U.S. Rollout TechCrunch
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. Lire la suite

Anthropic releases Claude Fable 5, turning single prompts into playable games and detailed maps

Anthropic releases Claude Fable 5, turning single prompts into playable games and detailed maps TechCrunch
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. Lire la suite