The latest news on OpenClaw, AI agents, and automation
Tag
108 articles
OpenAI introduced Guaranteed Capacity on May 19, letting enterprises lock in one to three-year compute commitments with volume discounts. CEO Sam Altman said the world will be 'capacity-constrained for some time,' positioning reserved compute access as the new competitive layer for companies scaling AI agents in production.
Peter Steinberger, the OpenClaw founder who joined OpenAI in February, disclosed that his three-person team running 100 Codex coding agents spent $1,305,088 on OpenAI API tokens in a single month. The figure puts a hard number on what agentic development at scale actually costs, and reveals just how heavily AI coding tools subsidize their subscription pricing.
OpenAI told staff on Friday that co-founder Greg Brockman will permanently lead all product strategy, merging ChatGPT, Codex, and the developer API into one organization. The restructuring formalizes what started as an interim arrangement during Fidji Simo's medical leave and signals OpenAI's commitment to a single agentic platform ahead of a potential late-2026 IPO.
OpenAI launched personal finance tools for ChatGPT Pro users on May 15, connecting bank accounts, credit cards, and investment portfolios from 12,000+ institutions through Plaid. The feature provides spending dashboards, portfolio tracking, and AI-driven financial planning, one month after OpenAI acquired personal finance startup Hiro.
OpenAI's Codex coding agent is now available inside the ChatGPT mobile app, letting developers remotely monitor tasks, approve commands, and steer autonomous coding sessions running on their Mac. The preview, available across all ChatGPT plans, turns phones into a remote control layer for desktop agent workflows.
OpenClaw's latest release hands the inner model loop for OpenAI agents to Codex app-server by default. ChatGPT subscribers get native thread state, dynamic tool search, and cleaner separation between reasoning and delivery, while OpenClaw retains control of channels, memory, cron, and the product layer around the turn.
OpenAI will continue paying Microsoft 20% of revenue through 2030, but total payments are now capped at $38 billion. The revised deal removes the AGI opt-out clause, makes Microsoft's IP license non-exclusive, and lets OpenAI serve enterprise customers through AWS, Google Cloud, and other providers.
OpenAI launched Daybreak on Monday, a cybersecurity platform that uses Codex as an agentic harness to scan codebases, triage vulnerabilities, and generate patches with audit-ready evidence. Partners include Cloudflare, Cisco, CrowdStrike, Palo Alto Networks, Oracle, and Akamai. The initiative directly competes with Anthropic's Project Glasswing.
OpenAI announced Monday it will grant the European Commission, EU institutions, and vetted European cybersecurity teams preview access to GPT-5.5-Cyber through a new OpenAI EU Cyber Action Plan. Anthropic has declined to grant similar access for its Mythos model. The EU Commission confirmed 'four or five' meetings with Anthropic but said discussions are 'not yet at the same stage' as with OpenAI.
OpenAI formally launched the OpenAI Deployment Company on Monday with more than $4 billion in initial investment from 19 partners. The company acquired AI consulting firm Tomoro to staff the unit with approximately 150 Forward Deployed Engineers from day one. Brookfield committed $500 million. TPG leads the investor group alongside Advent, Bain Capital, Goldman Sachs, SoftBank, McKinsey, and Capgemini.
Supply chain analyst Ming-Chi Kuo reports OpenAI has moved its AI agent phone mass production target to the first half of 2027, up from 2028. The device will use a custom MediaTek Dimensity 9600 built on TSMC's N2P node, dual AI processors, and an enhanced camera ISP for real-world agent perception. Combined 2027-2028 shipments could reach 30 million units.
OpenAI deployed Auto-Review mode for its Codex coding agent, replacing manual human approval at sandbox boundaries with review by a separate GPT-5.4 Thinking agent. Internal data shows 200x fewer interruptions compared to default approval mode, with 99.1% of escalated actions auto-approved. The system caught 90.3% of overeager agent actions, 99.3% of prompt injection attempts, and preserved full audit trails via OpenTelemetry export.
Anthropic has partnered with Blackstone, Goldman Sachs, and Hellman & Friedman to launch a $1.5 billion enterprise AI services firm that will embed Claude engineers directly inside businesses. OpenAI is pursuing a parallel structure with The Development Company, reportedly raising $4 billion from 19 investors at a $10 billion valuation. Both moves signal that frontier AI labs now see enterprise agent deployment, not model licensing, as the primary growth lever.
A malicious Hugging Face repository impersonating OpenAI's Privacy Filter project reached #1 on the platform's trending list and accumulated 244,000 downloads before removal. HiddenLayer researchers discovered the campaign on May 7, finding that the repository's loader.py delivered a Rust-based infostealer targeting browser credentials, cryptocurrency wallets, and SSH keys.
Meta is building a personalized AI assistant for everyday tasks. Google is developing a 24/7 Gemini-powered personal agent. Both moves come months after OpenClaw's viral adoption proved consumer demand for AI that acts, not just answers, and OpenAI's February hire of creator Peter Steinberger signaled the stakes.
OpenAI released GPT-5.5-Cyber to a limited set of vetted cybersecurity teams responsible for protecting critical infrastructure. The model is trained to be more permissive on security workflows like red teaming and penetration testing while maintaining safeguards against malicious use. The release follows Anthropic's Mythos debut by roughly one month and escalates the competition over AI-powered security automation for autonomous defense agents.
Saline Township, Michigan unanimously rejected a 700-acre data center for OpenAI's Stargate initiative in September 2025. Two months later, construction began anyway after the developer sued for exclusionary zoning and the township settled. The facility's 1.4-gigawatt power appetite equals 25% of DTE's peak capacity, and the legal playbook is already being replicated at Stargate sites in Texas, Ohio, and Wisconsin.
OpenAI shipped three agent-enabling features for Codex in April: computer use (see, click, type in apps), an in-app browser for context retrieval, and automated PR review. Independent testing on a real Python codebase shows Codex now matches Claude Code's autonomous debugging capabilities. The developer-agent arms race just got a lot tighter.
Analyst Ming-Chi Kuo reports OpenAI has accelerated its AI agent phone timeline from 2028 to first-half 2027 mass production. The device will run a customized MediaTek Dimensity 9600 built on TSMC's N2P node, with dual-NPU architecture and an ISP designed for real-world visual sensing. Combined 2027-2028 shipments could hit 30 million units.
Ten days after Sam Altman called Anthropic's restricted model distribution 'fear-based marketing,' OpenAI adopted the same framework for GPT-5.5-Cyber. The first ground-up retrain since GPT-4.5 is positioned as agent-first, but the real shift is in who gets access to frontier capabilities.
OpenAI's CEO told Stripe co-founder Patrick Collison that building an OpenClaw agent to manage his morning communication overload was one of his biggest 'magic AGI moments' in the field. He later rebuilt the tool with OpenAI's Codex model, signaling the company's internal pivot from chatbot to agent-first product strategy.
OpenAI has closed more than $4 billion from 19 investors for a new joint venture called 'The Deployment Company,' designed to help enterprises deploy its AI software at scale. The move mirrors Anthropic's simultaneous $1.5 billion partnership with Blackstone and Goldman Sachs, marking a new phase where AI labs compete by financing customer adoption.
OpenAI shipped Codex Pets, optional floating animated companions for its desktop coding agent on Windows and macOS. The pets display real-time task status, notify developers when Codex needs input, and can be custom-generated with the /hatch command. Eight built-in options ship by default. The feature is a consumer UX play for agent transparency.
Sam Altman announced on May 2 that ChatGPT subscribers can now authenticate directly into OpenClaw and run autonomous agents via GPT-5.4 for $23 per month. The move arrives exactly one month after Anthropic banned Claude subscription access from the same platform, citing unsustainable compute costs. Two companies looked at the same 3.2 million users and made opposite bets: OpenAI chose distribution, Anthropic chose margin protection. The divergence reveals a fundamental strategic split in how the two leading AI labs plan to monetize the agent era.
OpenAI CEO Sam Altman outlined three strategic focus areas on the Nothing But Tech podcast: accelerating scientific research, enabling 'automated startups' where one-person teams leverage AI agents and compute to build companies, and developing 'personal AGI' with robotics as the physical layer. He described a future where humans should not remain the physical actuators of AI tasks.
The Department of Defense announced Friday that SpaceX, OpenAI, Google, Nvidia, Reflection AI, Microsoft, and Amazon Web Services have signed contracts allowing their AI models to be deployed for 'any lawful use' in classified military environments. The Pentagon has requested $54 billion for autonomous weapons development alone.
Sam Altman announced GPT-5.5-Cyber, a specialized frontier model for cybersecurity, rolling out first to vetted defenders. The restricted release follows Anthropic's Mythos playbook and signals that gated access for security-critical AI models is becoming the industry default.
OpenAI published Symphony, an open-source spec and Elixir reference implementation that turns project management boards like Linear into control planes for Codex coding agents. Agents pull tasks from issue trackers, run in isolated workspaces, monitor CI, resolve merge conflicts, and prepare pull requests for human review. Internal teams saw landed pull requests increase 500% in three weeks.
OpenAI's frontier models including GPT-5.5 are now available on Amazon Bedrock alongside a new co-built Managed Agents service for enterprise agent deployment. The move comes one day after OpenAI ended its exclusive cloud arrangement with Microsoft.
The Atlantic reports the Trump administration has multiple legal levers to seize or regulate frontier AI labs, from Defense Production Act invocations to utility-style rate controls. Full nationalization is unlikely. But soft nationalization, where the government takes equity stakes, places officials on boards, and embeds engineers inside labs, is already happening. For enterprise buyers building on these APIs, the question is no longer theoretical.
OpenAI and Microsoft overhauled their partnership on April 27, ending the exclusive cloud licensing arrangement that defined AI infrastructure since 2019. Microsoft keeps a 20% revenue share through 2030, now capped, while OpenAI gains the ability to distribute across AWS, Google Cloud, and any other provider.
Customers Bank CEO Sam Sidhu, who let an AI clone deliver his earnings call remarks, has signed a multiyear deal embedding OpenAI engineers on-site to automate commercial lending, deposits, and payments. The $25.9 billion lender expects to cut loan closing timelines from 45 days to seven and push its efficiency ratio from 49% into the low 40s.
Apple supply chain analyst Ming-Chi Kuo reports that OpenAI is working with Qualcomm and MediaTek on custom smartphone processors, with Luxshare as the exclusive manufacturing partner. Mass production is targeted for 2028, with finalized chip specs expected by late 2026 or Q1 2027.
OpenAI's GPT-5.5 scores 82.7% on Terminal-Bench 2.0 and 58.6% on SWE-Bench Pro while using fewer tokens than its predecessor. Co-founder Greg Brockman calls it a step toward the company's 'super app' vision combining ChatGPT, Codex, and an AI browser into a single enterprise service.
OpenAI's Python Agents SDK now includes Sandbox agents that run in isolated containers with full filesystem access, shell execution, package management, port exposure, snapshots, and persistent memory. Eight sandbox providers are supported including E2B, Modal, Docker, Vercel, Cloudflare, Daytona, Runloop, and Blaxel, each with different isolation models and performance characteristics.
OpenAI's new Workspace Agents let enterprise teams build shared AI agents that execute multi-step workflows, run on schedules, operate in Slack, and persist across sessions. Powered by Codex, the agents represent OpenAI's shift from conversational assistant to autonomous task execution platform. Free through May 6, with credit-based pricing after.
OpenAI's Sora video generation tool goes dark on April 26, ending a product that peaked at one million users but hemorrhaged roughly $1 million per day in compute costs. The API stays online until September 24. The shutdown is part of a broader purge of 'side quests' that also claimed three senior executives in a single day: former CPO Kevin Weil, Sora architect Bill Peebles, and enterprise CTO Srinivas Narayanan.
OpenAI has locked in more than $20 billion in compute spending with wafer-scale chip maker Cerebras Systems over the next three years, taking an equity stake and providing a $1 billion infrastructure loan. Cerebras filed for a Nasdaq IPO on April 17 after withdrawing its 2024 attempt, targeting a valuation of roughly $35 billion.
OpenAI demonstrated its cybersecurity-focused model to federal agencies, state governments, and Five Eyes intelligence partners at a Washington D.C. event, briefing approximately 50 cyber defense practitioners on capabilities available through its Trusted Access program.
OpenAI released GPT-5.5 on Thursday, its newest flagship model, to all paid ChatGPT tiers and Codex. The model scores 82.7% on Terminal-Bench 2.0 for agentic coding, matches GPT-5.4 latency while using fewer tokens, and carries OpenAI's first 'High' cybersecurity risk classification. API access is not yet available.
OpenAI's next model, internally codenamed Spud, is its first fully retrained base since GPT-4.5. GPT-5.5 introduces a 256,000-token context window, native multimodal processing across text, images, audio, and video, and step-by-step agent tool execution for web browsing, code, and API calls.
OpenAI launched workspace agents in ChatGPT on April 22, evolving custom GPTs into Codex-powered team automation. Agents run in the cloud, handle multi-step workflows across tools like Slack and CRMs, and keep working when users are offline. Available in research preview for Business, Enterprise, and Edu plans, with credit-based pricing starting May 6.
CGI, a $76 billion IT consulting firm with 94,000 professionals, announced a strategic expansion of its OpenAI partnership on April 21. Tens of thousands of CGI engineers and consultants already use Codex for agentic AI workflows spanning legacy code modernization, code review, and vulnerability detection.
OpenAI shipped GPT-5-Codex on April 21, a specialized GPT-5 variant built for autonomous code generation, review automation, and agentic workflows. The model is the default for cloud-based Codex tasks and code review, and selectable in the Codex CLI and IDE extensions for local development.
OpenAI's ChatGPT and Codex services went down on April 20 with over 5,000 users reporting issues. The outage, which persisted for hours, highlights the fragility of agent workflows built on single-provider API dependencies.
OpenAI's CFO confirmed the company is killing Sora, its AI video tool that cost $1 million per day to run, to reallocate compute toward Spud, a new enterprise-focused model. Enterprise revenue has doubled from 20% to 40% of OpenAI's total since 2024. But Anthropic just passed OpenAI in annualized revenue at $30 billion, three senior executives departed in a single day, and both companies are projecting billions in losses. This is the story of how the company that defined consumer AI decided consumer AI was the wrong bet.
OpenAI acquired AI personal finance startup Hiro Finance in an acqui-hire announced April 13. Founder Ethan Bloch previously sold neobank Digit for roughly $230 million and built an OpenClaw autotrading agent called RoboBuffett. Hiro shuts down April 20 with all user data deleted by May 13.
OpenAI's updated Agents SDK gives developers native sandbox execution across seven providers, a model-native harness with filesystem tools, and durable execution with snapshotting. The update targets the gap between prototype agents and production deployments where security, file access, and long-running reliability matter.
Tools for Humanity announced the biggest World ID overhaul yet at an event in San Francisco on April 17, pitching iris-scan plus zero-knowledge-proof verification as the identity layer for a world where AI agents routinely transact on behalf of humans. Tinder, Zoom, DocuSign, and Shopify are the first enterprise integration partners. The system now has 40 million app users and nearly 18 million verified humans across six continents.
OpenAI launched GPT-Rosalind on April 16, the first model in a new Life Sciences series fine-tuned for biochemistry and genomics reasoning. The model scored a 0.751 pass rate on BixBench (bioinformatics), outperformed GPT-5.4 on six of eleven LABBench2 tasks, and ranked above the 95th percentile of human experts on RNA sequence-to-function prediction in a partnership with Dyno Therapeutics. Access is gated through OpenAI's Trusted Access program, limited to US-based research organizations. For teams building drug discovery agents, genomics pipelines, or lab automation workflows, GPT-Rosalind is a purpose-built API endpoint to evaluate against general-purpose models.
OpenAI has agreed to pay chip startup Cerebras more than $20 billion over three years for servers powered by its wafer-scale chips, doubling a previous $10 billion commitment. The deal includes warrants for up to a 10% equity stake in Cerebras and $1 billion in data center funding, as OpenAI moves to reduce its reliance on Nvidia.
OpenAI's Codex can now control macOS desktop apps with its own cursor, run multiple agents in parallel without interrupting the user, schedule its own future work across days or weeks, generate images, and browse the web natively. The update shipped to Codex's 3 million weekly users on April 16.
The maker of Wegovy and Ozempic is integrating OpenAI's technology across its entire operation, from analyzing drug candidates to optimizing the manufacturing supply chain it has struggled to scale. Pilot programs launch immediately in R&D, manufacturing, and commercial operations, with full integration by year-end. CEO Mike Doustdar says the partnership enables Novo Nordisk to 'analyse datasets at a scale that was previously impossible.'
OpenAI has joined the FIDO Alliance, the standards body behind passkeys and biometric authentication for Apple, Google, and Microsoft devices, to shape how phones authenticate AI agents acting autonomously on users' behalf. 'We've joined FIDO to participate in emerging work to evolve authentication for agentic intelligence,' OpenAI stated.
OpenAI shipped a major Agents SDK update on April 15 with native sandbox execution across seven providers (including Cloudflare, E2B, Modal, and Daytona), a model-native harness with Codex-like filesystem tools, and a Manifest abstraction for portable workspaces. The update makes sandboxed agents the default, not an add-on, and lets developers bring their own compute infrastructure without vendor lock-in.
OpenAI released a cybersecurity-specialized variant of GPT-5.4 with lower refusal limits and binary reverse engineering as a model-native capability. Access is restricted to verified security professionals through a tiered identity verification program now expanded to thousands of defenders.
Security researchers at CALIF gave OpenAI's Codex a browser-level shell on a Samsung Smart TV and the matching firmware source code. Without being pointed at any specific driver or vulnerability, Codex enumerated the attack surface, audited Samsung's vendor kernel code, discovered world-writable memory-management interfaces, and built a privilege escalation chain to root. The full exploit ran against a live device running firmware shipped on millions of consumer TVs.
OpenAI has acquired Hiro Finance, a personal finance AI startup backed by Ribbit Capital and General Catalyst. Founder Ethan Bloch, who built an autonomous OpenClaw stock-trading agent called RoboBuffett and previously sold neobank Digit for over $200 million, joins OpenAI with roughly 10 employees to build AI financial tools. Hiro shuts down April 20.
OpenAI released ChatGPT agent, a unified system that moves ChatGPT from conversational AI to autonomous task execution. The agent can browse websites, write and run code, complete forms, manage calendars, and execute multi-step workflows with limited human involvement. It scored 41.6% on Humanity's Last Exam and 27.4% on Frontier Math, suggesting strong but imperfect capabilities that IBM researchers say raise new questions about oversight and control delegation.
OpenAI has put its Stargate UK infrastructure project on hold, pointing to industrial energy prices among the world's highest and unresolved AI copyright rules. The pause affects up to 8,000 Nvidia GPUs planned for Cobalt Park in North Tyneside and deals a blow to the UK government's AI growth strategy.
OpenAI announced a new $100 per month ChatGPT Pro subscription tier on April 9, offering 5x higher Codex usage limits than the $20 Plus plan. The tier targets heavy Codex users who need longer coding sessions but don't require the full $200 Pro plan. Codex now has three million weekly users, with Claude Code run-rate revenue exceeding $2.5 billion.
OpenAI is finalizing a cybersecurity product for restricted release to select partners, according to an Axios scoop published April 9. The product is separate from ChatGPT and follows OpenAI's existing Trusted Access for Cyber pilot program, which launched in February after GPT-5.3-Codex and provides vetted organizations with permissive models for defensive security research, backed by $10 million in API credits. The move mirrors Anthropic's restricted rollout of Claude Mythos Preview.
OpenAI sent letters to the California and Delaware attorneys general on April 6 asking them to investigate Musk's alleged anti-competitive behavior, weeks before the April 27 trial begins. On the same day, The New Yorker published a 100-source investigation concluding that OpenAI insiders don't trust Sam Altman. For agent builders, both stories point at the same risk: the dominant infrastructure layer under your stack is run by a company in institutional crisis at the exact moment it's commanding record valuations.
The New York Times published a major cybersecurity feature on April 6 confirming what agent builders already suspected: AI systems from Anthropic and OpenAI have fundamentally changed both the attack and defense sides of cybersecurity. Hackers are using AI agents to probe and breach infrastructure at a speed and scale previously impossible. The response from defenders is to deploy more AI. The result is an AI-versus-AI arms race running through every enterprise environment that relies on automated systems.
OpenAI published a policy paper on April 6 outlining a vision for managing AI's economic impact: robot taxes to shift the burden from labor to capital, a Public Wealth Fund to give citizens automatic stakes in AI infrastructure, and a subsidized four-day workweek. The paper's real signal for agent builders is what OpenAI assumes as a baseline: that autonomous AI systems are already disrupting labor markets at scale, and that redistribution mechanisms are necessary as a result.
Confidential financial documents from OpenAI and Anthropic, reviewed by the Wall Street Journal ahead of their funding rounds, show both companies face the same core problem: training costs are growing faster than revenue. OpenAI expects to spend $121 billion on compute by 2028 and won't break even until after 2030. Anthropic projects a break-even point sooner, but faces the same structural pressure. For builders running agents on either API stack, this is the financial context behind every pricing and policy decision both companies make.
Three companies that compete fiercely on pricing and capability have quietly begun sharing threat intelligence through the Frontier Model Forum to detect Chinese adversarial distillation: the technique of using API query outputs to train cheaper imitation models. Bloomberg reported the cooperation on April 6. The models being protected are the same ones powering agent harnesses built on Claude, Gemini, and GPT — and an eroded model moat means an eroded foundation for any agent stack built on top.
Anthropic, OpenAI, Microsoft, and Google have all published guidance recommending single-agent architectures first. Most builders default to multi-agent swarms anyway. That gap is how microservices debt gets rebuilt, one orchestration layer at a time — except this time it's hiding in hand-offs, prompt arbitration, and agent state instead of the network.
OpenAI's CFO told colleagues the company isn't ready for a 2026 IPO, citing $200 billion in projected cash burn and $600 billion in cloud commitments where key investors are also key suppliers. It's the third executive friction signal at OpenAI in three weeks — and the spending commitments at stake directly shape how much compute is available for agent workloads on OpenAI's platform.
OpenAI shares have become nearly unsellable on secondary markets, with $600 million in institutional shares finding zero buyers. Meanwhile, secondary marketplaces are registering more than $1.6 billion in demand for Anthropic shares at valuations 50% above its last funding round. Goldman Sachs is waiving carry fees on OpenAI shares while charging full carry on Anthropic. For agent builders choosing a primary model provider, the capital flow is a leading indicator of platform durability.
A Jacobin investigation traces a direct line from OpenAI's January 2024 removal of its 'military and warfare' usage ban, through a hiring spree of more than a dozen national security insiders, to a $200 million defense contract secured within hours of the Trump administration blacklisting Anthropic for refusing military use cases. For builders choosing which platform to build agents on, the divergence is now structural.
Iran's Islamic Revolutionary Guard Corps released a video on April 4 featuring satellite imagery pinpointing the location of OpenAI's 1-gigawatt Stargate AI datacenter in Abu Dhabi, threatening 'complete and utter annihilation' of the facility. The threat follows actual strikes on Oracle's Dubai office and AWS data centers in the region, escalating from general warnings issued on April 1 to specific targeting of AI compute infrastructure that powers GPT-based agent services worldwide.
OpenAI CEO of AGI deployment Fidji Simo is taking several weeks of medical leave for a neuroimmune condition, COO Brad Lightcap is moving to a 'special projects' role reporting to Sam Altman, and CMO Kate Rouch is stepping down to focus on cancer recovery. Greg Brockman will oversee product in Simo's absence. The reshuffle comes as OpenAI eyes a potential IPO this year and integrates its $3 billion OpenClaw acquisition.
OpenAI has acquired TBPN, a daily tech talk show popular among Silicon Valley circles, in what multiple outlets are calling its biggest media move yet. The deal, first reported by the Wall Street Journal, comes as OpenAI battles a QuitGPT movement, a Pentagon contract backlash, and Anthropic's Claude surging in downloads. For developers building on OpenAI's APIs, the company now owns a media channel that covers the same platform they depend on.
OpenAI has signed Helsinki-based Smartly as its first creative ad-tech partner, with the goal of building interactive, conversational ad formats inside ChatGPT. Unlike the static contextual placements launched in February, the new formats will let users click an ad and enter a chatbot-style dialogue with a brand. Smartly cited a Boots UK pilot on Meta — where conversational ads drove nearly 5x the sales of standard placements — as the template. The partnership arrives six weeks after OpenAI's ad pilot crossed $100 million in annualized revenue with more than 600 advertisers.
SoftBank Group executed the first $10 billion installment of its $30 billion OpenAI commitment through Vision Fund 2 on April 2, 2026. The capital is structured in three tranches arriving April, July, and October, and is earmarked for AI-native infrastructure: compute, data pipelines, and distribution. This is the first time committed capital from OpenAI's record $122 billion raise has moved from paper to deployment at this scale.
Bloomberg reports that roughly $6 billion in OpenAI shares can't find buyers on the secondary market, while investors have $20 billion in cash ready to deploy into Anthropic. Morgan Stanley and Goldman Sachs have dropped their carry fees on OpenAI share sales — a move typically reserved for assets with weakening demand. The divergence signals a market-level reassessment of which AI platform will dominate the agent infrastructure layer.
OpenAI has officially closed its record-breaking funding round at $122 billion, valuing the company at $852 billion. The final close adds retail investor participation ($3 billion via bank channels), ARK Invest ETF inclusion, and a $4.7 billion undrawn credit facility. The company now generates $2 billion per month in revenue, processes 15 billion API tokens per minute, and says an IPO is imminent.
OpenAI patched a DNS exfiltration technique in ChatGPT's code execution runtime on February 20, 2026. Check Point Research demonstrated how attackers could encode conversation data, uploaded files, and AI summaries into DNS queries to bypass network restrictions. The vulnerability exploited an asymmetry in sandbox design: direct HTTP calls were blocked, but DNS resolution remained unrestricted. The issue is closed. The architectural pattern matters for any team running code in sandboxed environments.
Speaking at ClawCon in Tokyo on Monday, Peter Steinberger gave his most substantive public interview since OpenAI hired him in February. He told AFP that 2026 will be 'the year of the general agent,' said big companies couldn't have built OpenClaw because they 'would have worried too much about what could go wrong,' and acknowledged security concerns about a 'cottage industry' of companies making the tool easier to install than he intended.
Phantom Labs, BeyondTrust's research arm, disclosed a command-injection vulnerability in OpenAI's Codex coding agent that allowed attackers to extract short-lived GitHub OAuth tokens by manipulating branch names during task creation. The flaw extended across Codex's web interface, CLI, SDK, and IDE integrations, and could have been weaponized at scale by embedding malicious payloads in GitHub branch names. OpenAI has patched the issue.
In a pre-IPO investor document viewed by CNBC, OpenAI disclosed that Microsoft provides 'a substantial portion of our financing and compute' and that failing to diversify partners could adversely affect the business. For the thousands of companies building autonomous agents on GPT APIs, the disclosure puts a name on a risk most have been quietly ignoring: single-vendor compute dependency in their foundation model provider.
OpenAI signed a lease for a 202,400-square-foot industrial warehouse at Richmond's Portside Commerce Center, county records show. The facility — previously vacant after battery startup Moxion Power collapsed — aligns with Business Insider's January report that OpenAI told employees it planned to open a second robotics lab in Richmond. The deal adds to over 1.6 million square feet of Bay Area space OpenAI has secured in recent months.
OpenAI has updated Codex with a searchable plugin directory, MCP server integrations, and shareable skills bundles — the same stack that Anthropic's Claude Code has offered since late 2025. All three major coding agents now speak MCP, and the competitive surface is shifting from model quality to ecosystem breadth.
OpenAI is discontinuing Sora, its AI video generation product, after a $1 billion Disney licensing deal fell apart. The shutdown marks another in a series of product pivots by OpenAI and raises questions about the company's ability to execute beyond language models.
SoftBank has secured a $40 billion unsecured bridge loan to fund additional investments in OpenAI, maturing in March 2027. The facility extends SoftBank's already massive $30 billion commitment to OpenAI and signals continued concentration of capital in single AI model builders rather than diversified agent frameworks.
OpenAI's ChatGPT advertising pilot has hit $100M annualized run rate in six weeks. For agent builders on OpenAI's API, the question is direct: will ad optimization incentives reshape how the company prioritizes agentic features, and will ads eventually reach agent-powered applications?
On March 25, OpenAI made seven distinct announcements in a single day: shutting down Sora, losing the $1 billion Disney deal, handing off safety oversight, revealing a new model codenamed 'Spud,' closing a $10 billion raise, committing $1 billion through its Foundation, and killing the ChatGPT shopping feature. Taken together, these moves reveal a company stripping consumer-facing products to concentrate entirely on the agent and AGI mission.
OpenAI is pulling back from Instant Checkout, the feature that let users buy products directly inside ChatGPT, after low adoption and merchant onboarding difficulties. The company is shifting to a product discovery model powered by its Agentic Commerce Protocol, with retailers like Walmart, Shopify, and Etsy building their own apps within ChatGPT instead.
The OpenAI Foundation, the nonprofit that controls OpenAI and ChatGPT, announced it will grant out at least $1 billion over the next year across life sciences, jobs and economic impact, AI resilience, and community programs. The commitment is part of a previously announced $25 billion philanthropic pledge. The Foundation granted out just $7.6 million in 2024, making this a 130x increase in annual grantmaking. Key hires include OpenAI co-founder Wojciech Zaremba as Head of AI Resilience.
OpenAI CFO Sarah Friar confirmed on CNBC's Mad Money that the company has raised an additional $10 billion, bringing its total fundraise to 'north of $120 billion.' New investors include Microsoft, Andreessen Horowitz, D.E. Shaw Ventures, MGX, TPG, and T. Rowe Price. The extension comes roughly a month after OpenAI announced a $110 billion round led by Amazon's $50 billion commitment, at a $730 billion pre-money valuation.
OpenAI is offering private equity firms guaranteed minimum returns of 17.5% and early access to its newest models as both AI giants court buyout firms for joint ventures aimed at locking in enterprise customers ahead of expected IPOs.
OpenAI confirmed it will begin showing ads to all ChatGPT Free and Go users in the United States in the coming weeks, expanding beyond its initial pilot that has reached roughly 5% of mobile users. The ad program launched at approximately $60 CPM with test commitments of $200,000 to $250,000 per brand. Truist estimates OpenAI will generate under $1 billion in ad revenue this year, scaling to $30 billion by 2030.
CNBC's GTC week analysis argues OpenClaw is exposing a flaw in the AI investment thesis — if an open-source agent framework makes the underlying model interchangeable, the $1 trillion combined valuation of OpenAI and Anthropic faces a structural challenge. Forrester, Greylock, and Seaport Research weigh in on what commoditization means for the AI stack.
OpenAI will hire 3,500 people across engineering, research, sales, and a new 'technical ambassador' role, per the Financial Times. The push comes as Anthropic captures 70% of first-time enterprise AI buyers.
Snowflake confirmed targeted layoffs affecting technical writing and documentation teams, weeks after signing a $200 million multi-year deal to integrate OpenAI's GPT-5.2 into its data cloud. CEO Sridhar Ramaswamy framed the cuts as aligning teams with 'long-term strategy.'
A freedom of information request reveals the UK Department for Science, Innovation and Technology has held zero trials under its OpenAI memorandum of understanding, eight months after the deal was announced. The only deployment: ChatGPT access for Ministry of Justice civil servants.
Semafor's analysis of the Microsoft-OpenAI cloud dispute reveals a strategic picture beyond the lawsuit threat: Microsoft holds economic leverage over OpenAI that makes litigation unnecessary, and the real fight is about who controls the infrastructure layer for agentic AI workloads.
Microsoft is considering legal action against OpenAI over a $50 billion deal that makes AWS the exclusive third-party cloud provider for OpenAI's Frontier enterprise platform. The dispute centers on whether stateful agent environments fall outside Microsoft's contractual exclusivity over stateless API calls.
OpenAI will acquire Astral, the company behind Python's uv package manager and ruff linter. The Astral team will join Codex, which now has over 2 million weekly active users. This is OpenAI's second acquisition in March, following the OpenClaw deal, and both are pending regulatory approval.
OpenAI and Snowflake signed a $200 million multi-year deal to embed GPT-5.4 natively in Snowflake Cortex AI. Data engineers can now invoke frontier reasoning models through standard SQL queries without data leaving the Snowflake perimeter — turning every enterprise data team into a potential OpenAI customer.
Cursor's second-generation coding agent outperforms Claude Opus 4.6 while trailing GPT-5.4 on coding benchmarks. Meanwhile, OpenAI is reportedly in acquisition talks with Astral, the company behind the uv package manager and ruff linter used by most AI Python developers.
New data from Digit.fyi and Axios shows Anthropic's Claude now accounts for 24.4% of business AI usage, up from roughly 4% a year ago. OpenAI's enterprise share is declining as buyers cite Claude's reliability and safety posture as the deciding factor.
OpenAI released GPT-5.4 Mini and Nano — two small, fast models explicitly marketed as 'subagents' designed to execute tasks delegated by a primary AI. Available to free ChatGPT users, they mark the first time a major lab has shipped models purpose-built for multi-agent hierarchies rather than direct human interaction.
CNET confirmed during GTC 2026 coverage that OpenAI acquired OpenClaw after it went viral, while Meta picked up its social spin-off Moltbook. The two largest AI companies now control the core infrastructure and social layer of the agentic internet.
A Morgan Stanley research note forecasts a sharp AI capability breakthrough between April and June 2026, citing GPT-5.4's 83% score on the GDPVal benchmark — at or above human expert level on economically valuable tasks. The bank estimates $3 trillion in enterprise value could shift within 12 months.
The Department of Defense filed a formal rebuttal calling Anthropic's AI safety red lines an 'unacceptable risk to national security.' OpenAI filled the gap within weeks through an AWS classified-network deal. 150 retired federal judges and 30+ employees from rival labs now back Anthropic's legal fight. The AI industry's most consequential loyalty test is playing out in federal court.
OpenAI shipped Codex Security on March 6, two weeks after Anthropic launched Claude Code Security on February 21. Both use LLM reasoning instead of pattern matching to find vulnerabilities. Anthropic says Opus 4.6 discovered 500+ zero-days in open-source libraries during testing.