50 articles found
OpenAI plans to nearly double its workforce from 4,500 to 8,000 employees by the end of 2026, with new hires across product development, engineering, research, sales, and technical ambassador roles. The expansion is driven by competitive pressure from Anthropic and efforts to secure government contracts and enterprise deals with private equity firms.
From OpenAI combining ChatGPT with its browser and code tools to Google testing desktop Gemini integration, major AI companies are abandoning the multi-app strategy. This consolidation reveals a critical shift in how tech giants view AI's future—and suggests the experimental phase is definitively over.
OpenAI is building a fully automated AI researcher system, with plans to create an autonomous AI research intern by September 2024 as a precursor to a complete multi-agent system expected in 2028. The article also covers OpenAI's merger of ChatGPT, web browser, and coding tools into a single app, plus broader AI industry news including security concerns, energy costs, and regulatory developments.
OpenAI is redirecting its research efforts toward building a fully autonomous AI researcher system, with plans to develop an "AI research intern" by September 2024 that can tackle small research problems independently, followed by a multi-agent system by 2028. The initiative combines work on reasoning models, agents, and interpretability, building on recent advances like the Codex agent system, with chief scientist Jakub Pachocki stating the goal is to create systems capable of handling complex problems in mathematics, physics, life sciences, and policy that are too large for humans to solve alone.
OpenAI is developing a unified desktop application that combines ChatGPT, its browser, and Codex code generator into a single super app. The consolidation aims to streamline user experience and focus resources on one product, with the company planning to develop agentic AI capabilities that can autonomously perform tasks like writing software and analyzing data with minimal human oversight.
Google is testing a macOS version of its Gemini AI assistant app, which would compete directly with ChatGPT and Claude on desktop. The app includes a 'Desktop Intelligence' feature that allows Gemini to view screen content and access information from applications to provide contextual responses, similar to capabilities already available in competitor apps.
OpenAI has developed monitoring techniques using chain-of-thought analysis to detect misalignment in internal coding agents, examining real-world deployments to identify safety risks. The work focuses on strengthening AI safety safeguards by analyzing how AI systems behave when performing code generation tasks.
OpenAI is acquiring Astral, a company focused on Python developer tools, to accelerate growth of its Codex AI model. The acquisition aims to enhance OpenAI's capabilities in powering next-generation Python development tools.
The Pentagon is reportedly planning to have AI companies train specialized model versions on classified military data to support military operations, with OpenAI and xAI recently signing agreements with the Defense Department. The initiative would involve training in secure facilities where the Pentagon retains ownership of the classified training data, though experts warn of potential security risks if the same models are deployed across different clearance levels within the military.
The Pentagon is planning to establish secure environments where AI companies like OpenAI and xAI can train generative AI models directly on classified military data, a significant expansion beyond current use cases where models only answer questions about classified information. This development aims to create more accurate military-specific AI versions but presents security risks, particularly the potential for classified information to be inadvertently exposed to unauthorized Defense Department personnel or entities sharing the same model.
In just one week, we've seen two landmark defense contracts signal a fundamental shift in robotics and AI development. From Gecko Robotics' record Navy deal to OpenAI's controversial Pentagon partnership, military applications are no longer a side business—they're becoming the primary driver of innovation and investment.
OpenAI has launched GPT-5.4 mini and nano models, with the mini version now available to free and paid ChatGPT users. GPT-5.4 mini offers improved performance over GPT-5.0 mini in reasoning, multimodal understanding, and tool use while running more than twice as fast, while GPT-5.4 nano is available via API for cost-efficient tasks like data classification and extraction.
This technology news roundup covers several AI developments including OpenAI's controversial Pentagon partnership for military applications, xAI's Grok facing a CSAM lawsuit, China approving a brain-computer interface for commercial use, and Anthropic hiring weapons experts to prevent AI misuse. The article also highlights various AI-related stories ranging from deepfake porn markets to AI's impact on legal proceedings and military operations.
OpenAI released GPT-5.4 mini and nano, smaller and faster language models optimized for coding, reasoning, and subagent tasks. GPT-5.4 mini achieves 2x faster performance than its predecessor while approaching GPT-5.4's capabilities on benchmarks like SWE-Bench Pro, while nano provides the most cost-effective option for simpler classification and data extraction tasks.
OpenAI Japan announces the Japan Teen Safety Blueprint, introducing stronger age protections, parental controls, and well-being safeguards for teens using generative AI.
OpenAI has reached an agreement with the Pentagon to use its AI technology in classified military environments, raising questions about potential applications in the escalating conflict with Iran. The article explores how OpenAI's generative AI could be deployed for target analysis and prioritization, drone defense, and other military operations, marking the first major use of conversational AI for real-time combat decision-making.
OpenAI has been hit with another lawsuit. This time, Encyclopedia Britannica took legal action against OpenAI, accusing the company of copyright and trademark infringements, as first reported by Reute...
OpenAI's forthcoming "adult mode" will allow users to engage in lewd conversations with ChatGPT, but not use the chatbot to generate explicit images, audio or video. In response to reporting from The...
Researchers from nearly 1,000 experts, including Texas A&M University, created "Humanity's Last Exam," a 2,500-question benchmark designed to challenge advanced AI systems where traditional tests like MMLU no longer suffice. The exam covers specialized topics across mathematics, humanities, natural sciences, and ancient languages, with all questions that current AI models could solve removed to ensure lasting difficulty. Early results show even the most advanced systems like GPT-4o (2.7%), Claude 3.5 Sonnet (4.1%), and OpenAI's o1 (8%) score far below human expert levels, revealing significant gaps between AI and true expert-level knowledge.
A coalition of Big Tech companies is working on a more comprehensive solution to combat online scams. As first reported by Axios, Google, Microsoft, LinkedIn, Meta, Amazon, OpenAI, Adobe and Match Gro...
OpenAI plans to integrate its Sora video generation model directly into ChatGPT, according to The Information, aiming to boost user engagement and give the standalone Sora app a second life after declining interest. The integration could significantly increase inference costs for OpenAI, with the company projecting over $225 billion in inference spending between 2026 and 2030, and the company may implement a credit-based monetization system similar to the current Sora app.
Wayfair uses OpenAI models to improve ecommerce support and product catalog accuracy, automating ticket triage and enhancing millions of product attributes at scale.
OpenAI describes how it built an agent runtime using its Responses API, shell tools, and hosted containers to enable secure and scalable AI agents with file access, tool integration, and state management. The development demonstrates the evolution from language models to autonomous agents capable of executing tasks in controlled environments.
AI companies have been spending a lot of time in court arguing copyright cases over the past year and the latest plaintiff is Gracenote, the metadata company owned by Nielsen. Axios reports that Grace...
OpenAI is rolling out interactive visual responses in ChatGPT that generate dynamic, editable diagrams for over 70 math and science concepts like the Pythagorean theorem and Coulomb's law. Users can adjust variables and equations to see real-time effects on solutions, with the feature available to all ChatGPT users as part of OpenAI's expanded educational tools.
The article covers several AI-related developments including the Pentagon-Anthropic surveillance dispute, new White House AI rules requiring companies to allow "lawful" model uses, and the ongoing feud between OpenAI and Anthropic over DoD contracts. It also discusses Geoffrey Hinton stepping down from Google to focus on AI safety concerns, along with various other AI policy and capability stories.
OpenAI is acquiring Promptfoo, an AI security platform designed to help enterprises identify and remediate vulnerabilities in AI systems during development. The acquisition strengthens OpenAI's capabilities in AI safety and security testing infrastructure.
Caitlin Kalinowski's resignation from OpenAI over Pentagon partnership concerns represents more than one executive's protest—it's a canary in the coal mine for an industry racing ahead of its own ethical frameworks. When top robotics talent chooses principle over paycheck at AI's most influential company, we need to ask what they're seeing that boardrooms aren't.
Caitlin Kalinowski, OpenAI's robotics hardware lead, has resigned in protest over the company's hastily negotiated Department of Defense partnership, citing concerns about surveillance of Americans without judicial oversight and autonomous weapons development without proper guardrails. OpenAI stated it does not support domestic surveillance or autonomous weapons and believes the Pentagon agreement includes responsible safeguards, though CEO Sam Altman indicated willingness to amend terms to prohibit spying on Americans.
Here comes another disappointment for ChatGPT users. As first reported by Sources' Alex Heath, OpenAI is yet again delaying its "adult mode" for ChatGPT. A company spokesperson told Heath that "we're...
OpenAI's latest research reveals that reasoning models struggle to control their internal thought processes—and the company is celebrating this apparent weakness. This counterintuitive finding exposes a fundamental tension in AI development: the line between transparent reasoning and strategic deception.
Caitlin Kalinowski, OpenAI's head of robotics, resigned over concerns that the company rushed into a Department of Defense partnership without establishing proper guardrails against domestic surveillance and autonomous weapons. OpenAI disputed her concerns in a statement, claiming their DoD agreement includes red lines prohibiting domestic surveillance and autonomous weapons, while CEO Sam Altman indicated he would amend the deal to further prohibit spying on Americans.
MIT Technology Review's newsletter covers 10 major AI developments including Anthropic's lawsuit against the Pentagon over a software ban, the revelation that the Pentagon has been secretly testing OpenAI models for years, and various AI policy and corporate developments. The newsletter also highlights emerging research into 'AI societies' where hundreds of AI agents built their own civilization in Minecraft, and discusses AI's integration into smart homes and its energy consumption implications.
The UK government is delaying controversial AI copyright legislation that would have allowed AI companies like Google and OpenAI to train models on copyrighted materials without consent. Following a two-month consultation period with widespread opposition from creators, publishers, musicians, and filmmakers, the government has postponed the bill and will reconsider its approach, with no expectation it will appear in the May King's Speech.
Descript leverages OpenAI's models to enable multilingual video dubbing at scale, with optimizations that balance translation accuracy and timing to ensure dubbed speech sounds natural across different languages. The approach addresses the technical challenge of synchronizing translated dialogue with original video timing constraints.
OpenAI has released GPT-5.4, a new frontier model designed for professional work with native computer-use capabilities, improved coding and data analysis performance, and enhanced web research. The model features better desktop navigation, reduced factual errors (18% fewer than GPT-5.2), and will be the default model for ChatGPT's Thinking mode, available to enterprise customers and Pro subscribers.
OpenAI introduced CoT-Control, a new method for evaluating how well reasoning models can control their chains of thought during problem-solving. The research found that reasoning models struggle to maintain control over their internal reasoning processes, which OpenAI argues strengthens the case for monitorability as a key AI safety mechanism, as it makes model reasoning more observable and potentially more alignable with human oversight.
OpenAI has announced GPT-5.4, a new frontier AI model designed for professional applications featuring state-of-the-art coding capabilities, computer use functionality, tool integration, and extended 1M-token context window for processing longer inputs.
The White House announced a Ratepayer Protection Pledge with major tech companies including Amazon, Google, Meta, Microsoft, OpenAI, Oracle, and xAI to prevent AI data center electricity demands from raising consumer electricity bills. The pledge commits companies to funding new power generation, infrastructure upgrades, and separate rate structures, though it lacks binding enforcement mechanisms and doesn't address broader impacts on communities and critical computing resources.
OpenAI shares new tools, certifications, and measurement resources to help schools and universities close AI capability gaps and expand opportunity.
OpenAI introduces ChatGPT for Excel and new financial app integrations, powered by GPT-5.4 to accelerate modeling, research, and analysis in regulated environments.
OpenAI has released a Windows version of its Codex coding app, following its earlier macOS launch in February. The app enables users to coordinate multiple AI coding agents, includes automation features for tasks like bug testing, native Windows sandboxing, and a Skills section for connecting agents to specific tools and workflows.
OpenAI introduces the Learning Outcomes Measurement Suite to assess AI's impact on student learning across diverse educational environments over time.
OpenAI has reached a deal allowing the US military to use its technologies in classified settings, with CEO Sam Altman characterizing the negotiations as 'definitely rushed' and emphasizing safeguards against autonomous weapons and mass surveillance. The agreement represents a shift from Anthropic's refusal of similar Pentagon terms and raises questions about whether OpenAI can implement promised safety precautions amid military deployment pressures.
Meta has begun testing an AI shopping assistant feature on its Meta AI platform, allowing select US users to receive personalized product recommendations with pricing, images, and links to e-commerce sites. The feature, which uses user data like gender and location to tailor suggestions, represents Meta's entry into the competitive AI shopping assistant market alongside similar tools from OpenAI, Google, and Perplexity.
OpenAI's Sam Altman announced the company will amend its Defense Department agreement to explicitly prohibit the use of its AI systems for mass domestic surveillance of U.S. persons, adding specific language addressing Fourth Amendment protections and excluding use by intelligence agencies like the NSA without contract modification. The announcement comes after OpenAI faced backlash for quickly signing the deal following Trump's order to ban Anthropic services, while Anthropic refused similar demands from the Defense Department to remove AI safety guardrails for mass surveillance and autonomous weapons development.
On February 28, OpenAI announced it had reached a deal that will allow the US military to use its technologies in classified settings. CEO Sam Altman said the negotiations, which the company began pur...
A newsletter roundup covering anti-AI protests in London organized by activist groups, developments in AI policy including Pentagon negotiations with Anthropic and OpenAI, and upcoming AI model releases from DeepSeek. The coverage focuses on the growing public discourse around AI safety, energy consumption, and geopolitical competition in AI development.
From OpenAI's pledge to notify authorities of threats to Canadian demands for safety changes, a seismic shift is underway in AI governance. After years of operating in a regulatory gray zone, AI companies are discovering that 'move fast and break things' has real-world consequences—and governments worldwide are demanding answers when those things that break are public safety protocols.
Details on OpenAI's contract with the Department of War, outlining safety redlines, legal protections, and how AI systems will be deployed in classified environments.
