Actualités — Page33

OpenAI unveiled GPT-5.5, its newest language model powering ChatGPT, as the company pivots from pure conversation to autonomous, real‑world productivity. The upgrade rolls out across ChatGPT and Codex for Plus, Pro, Business and Enterprise tiers, with a premium “Pro” version for high‑level subscribers. Built to interpret loosely structured prompts, plan workflows and self‑check output, GPT-5.5 aims to cut back‑and‑forth interactions and deliver end‑to‑end results in coding, research, document creation and data analysis.Lire la suite

The White House Office of Science and Technology Policy released a memo on Wednesday alleging that entities in China are running industrial‑scale campaigns to distill U.S. artificial‑intelligence models. The memorandum pledges to share threat intelligence with American AI firms and to explore sanctions against the perpetrators. The claim builds on accusations from OpenAI and Anthropic that Chinese labs have used millions of queries to replicate frontier models. Lawmakers responded with the Deterring American AI Model Theft Act, while the memo arrives weeks before a planned Trump‑Xi summit in Beijing.Lire la suite

Generative artificial intelligence is amplifying the production of child sexual abuse material, prompting a sharp rise in reports to watchdogs and law‑enforcement agencies. Reuters found actionable AI‑generated CSAM reports more than doubled over two years, while the Internet Watch Foundation logged 8,029 such images and videos in 2025 alone. The National Center for Missing & Exploited Children received 1.5 million AI‑linked reports that year, up from 67,000 the previous year. A Minnesota case involving a school employee illustrated how everyday photos can be weaponized, leaving investigators to untangle whether a child in an image is real, altered or entirely fabricated.Lire la suite

A Wired investigation reveals that 11 gas‑powered AI data centers under construction or announced in the United States could release up to 129 million tons of carbon dioxide annually—more than Morocco’s total emissions in 2024. The facilities, linked to OpenAI, Meta, Microsoft and xAI, plan to run dedicated natural‑gas power plants to sidestep grid constraints, raising fresh concerns about the climate impact of the AI boom.Lire la suite

Anthropic announced Thursday that its Claude AI assistant now supports more than 200 third‑party connectors, adding a slate of consumer‑focused services such as Instacart, Spotify, Uber and TripAdvisor. The new integrations let users retrieve data and trigger actions directly from chat, while the system automatically suggests the most relevant connector based on conversation context. Anthropic says the feature remains ad‑free, that user data won’t be used to train models, and that users must confirm any major transaction before it completes.Lire la suite

OpenAI rolled out ChatGPT 5.5 this week, extending the new model to paying ChatGPT and Codex subscribers and promising an API release soon. Marketed as a work‑focused upgrade, the model excels at coding, computer‑task orchestration and research‑intensive workloads. President Greg Brockman highlighted its ability to interpret vague problems and act with minimal human direction, positioning it as a step toward true digital assistants. OpenAI also emphasized stronger cybersecurity safeguards, reflecting growing concerns about AI‑driven threats.Lire la suite

SpaceX announced Wednesday that it has entered a partnership with AI coding platform Cursor, giving the rocket firm access to the company’s agentic coding technology and a right to acquire Cursor later this year for up to $60 billion. The deal hinges on SpaceX’s Colossus supercomputer in Memphis, Tennessee, which boasts the equivalent of a million Nvidia H100 GPUs, and aims to create the world’s most capable coding and knowledge‑work AI. Both companies say the collaboration will accelerate model development and could eventually feed into Elon Musk’s broader X ecosystem.Lire la suite

OpenAI announced the rollout of its latest language model, GPT-5.5, on Thursday. The company describes the new system as its "smartest and most intuitive" model yet, capable of handling complex, multi‑step tasks such as writing and debugging code, conducting online research, and generating spreadsheets across a range of tools. GPT-5.5 uses fewer tokens, offers its strongest safety measures to date, and will be available to Plus, Pro, Business and Enterprise ChatGPT users as well as Codex subscribers. The launch intensifies the rivalry with Anthropic and comes just days before a high‑profile trial involving OpenAI executives.Lire la suite

Anthropic disclosed that a small group of unauthorized users gained access to its newly released Claude Mythos model on the day the company announced a limited rollout. According to Bloomberg, the intruders guessed the model’s online location using details leaked from a prior breach at data‑training firm Mercur and insider knowledge from a contractor who had evaluated Anthropic’s models. Anthropic said it is investigating the incident and reviewing its monitoring systems, which were designed to log and track model usage. The breach, described by security researchers as a standard “educated guess” attack rather than a sophisticated exploit, did not appear to target the model’s advertised cybersecurity capabilities. The episode raises questions about the robustness of Anthropic’s security controls for a product it has marketed as a “watershed moment” for defending digital infrastructure.Lire la suite

OpenAI unveiled Images 2.0, its latest AI image generator that can reason about composition, search the web for context, and produce up to eight coherent images from a single prompt. The model claims 99% accuracy in rendering text across multiple scripts and surged to the top of the Image Arena leaderboard within 12 hours, outpacing Google’s Nano Banana 2 by a record margin. Free‑tier ChatGPT users receive a basic version, while paid subscribers unlock the full "thinking" mode that includes web‑search and batch generation. Pricing starts at $0.04 per image, with higher‑resolution outputs up to 2K.Lire la suite

Google unveiled Workspace Intelligence, a new feature that integrates its Gemini AI model across the Google Workspace suite. The tool can pull data from Gmail, Docs, Sheets, Slides and Drive, using personal context to draft emails, edit documents, generate presentations and summarize conversations without users having to supply extra details. By tapping into stored emails, chats and files, Gemini promises to streamline routine tasks and keep output consistent with company style guides. The rollout follows Google’s earlier Personal Intelligence rollout, which let Gemini access Gmail and Photos for personal use.Lire la suite

OpenAI announced the debut of workspace agents, a new class of shared AI assistants that can plan, execute, and monitor multi‑step tasks for entire teams. Running in the cloud, the agents can access files, run code, and integrate with tools like Slack, freeing users from repetitive work and keeping projects moving even when they step away. The feature marks a shift from ChatGPT as a question‑answering bot toward a continuous work partner, though it still requires human oversight and proper permissions.Lire la suite

Google is weaving its Gemini AI assistant directly into Google Chat, turning the messaging platform into a one‑stop hub for routine work. Users can type natural‑language commands to draft Google Docs, build slide decks, pull files and book meetings without leaving the chat thread. The rollout also brings a daily briefing that surfaces urgent items and integrates third‑party tools such as Asana, Jira and Salesforce. By merging communication with execution, Google aims to cut the back‑and‑forth between apps and sharpen the productivity edge of Workspace.Lire la suite

OpenAI announced a partnership with Indian IT giant Infosys to integrate its artificial‑intelligence suite, including the Codex coding assistant, into Infosys' Topaz AI platform. The collaboration aims to help Infosys' global clients modernize software development, automate workflows and scale AI deployments, starting with software engineering, legacy modernization and DevOps. The move comes as IT services firms grapple with slowing client spending and rapid advances in generative AI, and offers OpenAI a distribution channel to reach large enterprises across more than 60 countries.Lire la suite

At the Google Cloud Next conference, the company unveiled a suite of AI-driven tools for its Workspace productivity platform. Named Workspace Intelligence, the new features let users tap Gemini to draft emails, generate documents, and auto‑populate spreadsheets. Administrators retain control over data access, and Google promises up to nine‑times faster data entry in Sheets. The rollout marks Google’s push to embed generative AI deeper into everyday office workflows, positioning its entrenched suite against rivals from Microsoft to emerging startups.Lire la suite

Anthropic confirmed it is investigating a report that a group gained unauthorized entry to its Claude Mythos model through a third‑party vendor portal. The breach, discovered via internet‑sleuthing tools and a developer portal, appears limited to exploratory testing rather than malicious exploitation. Anthropic’s Claude Mythos, released under the Project Glasswing preview, had been limited to a handful of trusted firms such as Amazon, Microsoft, Apple, Cisco and Mozilla, which used the model to identify hundreds of software flaws. The incident has revived concerns about AI‑driven cyber threats and the company’s recent designation as a supply‑chain risk by the U.S. Department of Defense.Lire la suite

Anthropic caused a stir on Monday when new subscribers to its $20‑per‑month Pro plan could not access Claude Code, the company’s agentic development tool. The move, later clarified as a test on roughly 2% of new sign‑ups, prompted backlash on Reddit and X. Anthropic’s head of growth, Amol Avasare, said the experiment reflected shifting usage patterns and that the feature would soon return to the Pro tier.Lire la suite

Anthropic’s new AI‑driven security tool, Mythos Preview, is being tested by several U.S. federal agencies, but the Cybersecurity and Infrastructure Security Agency (CISA) reportedly lacks access. While the Commerce Department and the National Security Agency are evaluating the model, CISA’s exclusion comes amid broader budget cuts and staffing limits imposed by the Trump administration, prompting questions about the nation’s readiness to defend critical infrastructure.Lire la suite

OpenAI is rolling out cloud‑based “workspace” agents to users of its Business, Enterprise, Edu and Teachers plans. The agents, built into ChatGPT, can automate tasks such as gathering product feedback and drafting follow‑up emails, and they integrate with tools like Slack and Gmail. OpenAI says the agents learn from team processes, request approvals when needed, and can be shared across an organization. The feature expands on the company’s 2023 GPT custom‑chatbot offering and comes as competition from Anthropic and other AI firms intensifies.Lire la suite

A small group of users gained entry to Anthropic’s restricted Claude Mythos Preview AI model on the day the company announced its launch, exploiting a third‑party vendor environment by guessing the model’s URL. Anthropic confirmed it is investigating the incident and said there is no evidence the breach affected its core systems. The episode highlights vulnerabilities in the way frontier AI tools are shielded behind external partners, raising concerns about the security of powerful cybersecurity AI models that can autonomously discover and exploit zero‑day vulnerabilities.Lire la suite