Actualités — Page34

Palo Alto‑based NeoCognition, a spin‑out from Ohio State University, announced a $40 million seed round led by Cambium Capital and Walden Catalyst Ventures. The startup aims to close the reliability gap in current AI agents by giving them the ability to build domain‑specific world models through on‑the‑job learning. Backed by investors such as Vista Equity Partners, Intel CEO Lip‑Bu Tan and Databricks co‑founder Ion Stoica, NeoCognition plans to target enterprise SaaS vendors with agents that improve performance within each software’s unique environment.Lire la suite

At Cloud Next 2026, Google announced a sweeping rebrand of its AI offerings, merging Vertex AI, Agentspace and new services into the Gemini Enterprise Agent Platform. The rollout includes Workspace Studio, a no‑code builder for AI agents across Gmail, Docs and other Google apps, a Model Garden with more than 200 models, and the production‑grade Agent2Agent protocol now in use at 150 enterprises. CEO Thomas Kurian framed the move as owning the entire stack—from custom TPU silicon to the inbox—positioning Google against rivals that, he said, “hand you the pieces, not the platform.”Lire la suite

OpenAI rolled out ChatGPT Images 2.0 today, a new image‑generation system that adds reasoning steps before creating visuals. The upgrade handles complex prompts, keeps concepts consistent across variations, and markedly improves text placement inside images—an area where earlier models struggled. The feature is available to all ChatGPT and Codex users, with advanced capabilities unlocked for Plus, Pro, Business and Enterprise plans, and the underlying model, gpt-image-2, is now offered via the API.Lire la suite

SpaceX announced a partnership with software‑development AI startup Cursor to build a next‑generation coding assistant. The agreement includes an option for SpaceX to buy Cursor for up to $60 billion later this year, or to pay $10 billion for its services. The deal leverages SpaceX’s Colossus supercomputer and ties into the broader Musk‑led ecosystem that already includes xAI and the X social platform.Lire la suite

A private online forum has reportedly breached Anthropic’s newly unveiled cybersecurity AI, Mythos, according to Bloomberg. The group, linked to a Discord channel that hunts unreleased AI models, accessed the tool through a third‑party contractor that works with Anthropic. Anthropic confirmed it is investigating the incident but said no evidence yet shows the breach affected its own systems. Mythos, rolled out to a handful of vendors such as Apple under the Project Glasswing initiative, was designed to strengthen enterprise security, raising concerns that the tool could be repurposed by malicious actors.Lire la suite

OpenAI announced a major upgrade to its ChatGPT image generator, unveiling ChatGPT Images 2.0 in a livestream briefing. The new model introduces a reasoning phase that lets the system parse complex prompts before creating visuals, resulting in more accurate text rendering, consistent styles and better layout control. By treating prompts as instructions rather than suggestions, the update narrows the gap with rival Google Gemini and promises fewer retries for users seeking polished graphics. CEO Sam Altman hailed the leap as a shift comparable to moving from GPT‑3 to GPT‑5 in a single step.Lire la suite

VPN provider Windscribe has rolled out native support for OpenClaw, an agentic AI platform that runs on local hardware. The new integration lets autonomous AI agents modify VPN settings, routing their traffic through encrypted tunnels and shielding users' home IP addresses. Windscribe says the feature addresses a privacy blind spot for AI‑driven bots, often called “lobsters,” that can expose real locations when they make web requests. The open‑source skill works with any compatible AI framework and is available on both paid and free Windscribe plans.Lire la suite

OpenAI announced Tuesday that its new ChatGPT Images 2 model is now live for all users. Built to excel at text‑heavy designs such as infographics, study guides and marketing graphics, the service expands the visual capabilities of the ChatGPT platform. The model produces clearer typography, supports multiple languages and offers 2K‑ and 4K‑resolution outputs through the API, though the higher resolutions remain in beta. Pricing follows existing subscription tiers, with higher‑paying plans receiving larger generation limits. OpenAI also continues to embed C2PA metadata and enforce strict content‑policy safeguards.Lire la suite

AI research startup NeoCognition announced a $40 million seed financing led by Cambium Capital and Walden Catalyst Ventures, with participation from Vista Equity Partners, Intel CEO Lip‑Bu Tan and Databricks co‑founder Ion Stoica. The Ohio State‑based lab, headed by professor Yu Su, aims to create agents that can autonomously specialize across domains, addressing the reliability gap that plagues current AI assistants. The company plans to market its self‑learning agent platform to enterprise SaaS providers, leveraging Vista’s extensive software portfolio.Lire la suite

OpenAI rolled out ChatGPT Images 2.0 on April 21, 2026, offering users the ability to generate multiple, higher‑resolution images from a single prompt and render readable text within those visuals. The new model taps into ChatGPT’s reasoning engine, supports custom aspect ratios, and extends its knowledge base to December 2025. While English text rendering shows marked improvement, the system still struggles with non‑English languages. The upgrade is available to all ChatGPT and Codex users, with a more powerful tier for paid subscribers.Lire la suite

OpenAI chief Sam Altman accused rival Anthropic of using fear‑mongering to promote its new cybersecurity model, Mythos, during a recent podcast appearance. Altman suggested the rhetoric was designed to keep advanced AI tools in the hands of a select few, echoing broader industry debates about hype, safety and market positioning.Lire la suite

OpenAI has rolled out ChatGPT Images 2.0, a new image‑generation model that promises sharper detail, wider aspect ratios and a marked leap in handling non‑Latin scripts. Available today to all ChatGPT users, the upgrade offers up to 2K resolution, flexible output formats and a reasoning layer that can verify its own results. The company says the model now produces more accurate depictions of Japanese, Korean, Chinese, Hindi and Bengali text, making it a stronger tool for developers, designers and creators who need reliable visual content.Lire la suite

OpenAI unveiled ChatGPT Images 2.0, a revamped AI image generator that can browse the web for reference, produce up to eight coordinated images per prompt, and output pictures up to 2K resolution. The upgrade, powered by the new GPT‑Image‑2 model, adds “thinking” capabilities for better detail retention, multilingual text rendering and a broader range of aspect ratios. The features roll out to ChatGPT Plus, Pro, Business and Enterprise users, while all ChatGPT accounts receive core improvements to photo realism and style fidelity.Lire la suite

Florida Attorney General James Uthmeier announced that his Office of Statewide Prosecution has opened a criminal investigation into OpenAI and its ChatGPT service after a suspect allegedly used the AI tool while planning the 2025 Florida State University mass shooting. The inquiry will examine whether the chatbot’s responses constitute aiding or abetting a crime under state law. OpenAI says the model provided only factual information and has cooperated with investigators, sharing account data and policy documents. The case marks the first time a U.S. state has pursued criminal liability against an artificial‑intelligence provider for a violent act.Lire la suite

President Donald Trump told CNBC that a deal allowing Anthropic’s AI models to be used by the Department of Defense is "possible," reversing a February order that barred federal agencies from the firm’s technology. The comment followed a White House meeting on April 18 in which Anthropic CEO Dario Amodei discussed the company’s new Mythos model with Chief of Staff Susie Wiles and Treasury Secretary Scott Bessent. Anthropic remains under a supply‑chain‑risk designation that bars it from Pentagon contracts while two federal courts issue conflicting rulings on the ban.Lire la suite

Amazon has deepened its partnership with Anthropic, pouring an extra $5 billion into the AI startup. The cash boost brings Amazon's total immediate stake to $13 billion and opens the door for Anthropic to secure up to 5 gigawatts of Amazon‑built AI chips by 2026. The infusion arrives as Anthropic grapples with soaring demand for its Claude models, which has strained its cloud infrastructure and caused occasional outages. The deal also includes a potential $20 billion follow‑on investment if the collaboration meets certain commercial milestones.Lire la suite

Educators and tech observers say AI‑generated text is flooding the internet, and the tools meant to flag it are falling short. Professors report a surge in perfectly grammatical but soulless writing that mimics human style only on the surface. The rise of ChatGPT, Claude and similar models has left schools scrambling for reliable ways to identify machine‑written work, as existing detectors struggle to keep pace with ever‑more sophisticated outputs.Lire la suite

Google has assembled an internal “strike team” to accelerate the development of its Gemini AI agents after co‑founder Sergey Brin warned that the company is falling behind Anthropic’s Claude Code. The effort, overseen by Brin and DeepMind CTO Koray Kavukcuoglu, aims to boost Gemini’s ability to write and execute code, a capability Brin says is essential for the final sprint toward AI takeoff. While the team’s work may initially remain internal, Google hopes the improvements will eventually translate into more powerful models for both its engineers and, potentially, the broader market.Lire la suite

Amazon announced a fresh commitment of $5 billion to AI start‑up Anthropic, with the option to add another $20 billion if the firm hits predefined milestones. The deal builds on two prior $4 billion rounds and ties Anthropic to Amazon’s custom Trainium chips and a decade‑long $100 billion spend on AWS services. In return, Anthropic will make its Claude platform available directly through the AWS console, eliminating separate login steps for customers.Lire la suite

Amazon has poured an additional $5 billion into AI startup Anthropic, raising its total stake to $13 billion. In return, Anthropic pledged to spend more than $100 billion on Amazon Web Services over the next decade, gaining access to up to 5 GW of new computing capacity and a suite of Amazon’s custom AI chips. The deal mirrors a similar arrangement Amazon struck with OpenAI two months earlier and underscores the tech giant’s push to lock in next‑generation generative‑AI workloads on its cloud platform.Lire la suite