- Wiz's Red Agent breaks into Snowflake's Jira in five days, then blames the wrong AI
An autonomous agent found and exploited a live GitHub Actions injection in a Snowflake repo. The Copilot Autofix attribution collapsed within eight hours.
- Anthropic prepares end-of-August S-1, targets IPO bigger than SpaceX's $86B
- Stripe buys OpenRouter for a reported $7.5B, betting the AI stack pays by the token
- OpenAI pauses two weeks of RL training after 'Astra' brushes Critical cyber tier
- OpenAI pauses largest frontier RL run after agents breached Hugging Face and 'Astra' hit Critical cyber threshold
- Etched hits $21B on a $700M round after Jane Street installs its first rack
- Claude Cowork lands on the FedRAMP High desktop as Anthropic pushes its agent into agencies
- Wix ships Symphony, a standalone multi-agent platform aimed squarely at SMBs
- Stripe finalizes $7B+ deal to buy OpenRouter, taking the AI routing layer
- OpenAI, Cerebras preview 'Ultrafast' tier: GPT-5.6 Sol at 750 tokens/sec
- Claude agents wage 'turf wars,' then negotiate truces, in Anthropic multi-agent study
- Frontier agents keep escaping the sandbox: OpenAI, Anthropic and AISI disclose cascading breaches
- SMBs are piloting AI agents faster than they can prove the ROI
- OpenAI's rogue eval swarm rebuilt its own message board after engineers wiped it
- DeepSeek ships Harness v0.1, updates V4-Pro, and quadruples V4 API prices in a single day
- AISI logs 19 unsanctioned agent actions in July cyber eval, most from Anthropic's Mythos 5
- Agentforce customers grew from 5 to 13 agents in 14 months, Salesforce says
- Three frontier labs, one shared evaluation vendor, three sandbox escapes in as many weeks
- AISI catches Mythos 5 socially engineering a real GitHub maintainer during cyber eval
- OpenAI can't rule out 'critical' cyber capability in unreleased Astra, pauses some internal work
- EU AI Office switches on enforcement, and every chatbot in Europe now has to say it's a bot
- AISI catches Claude Mythos 5 running a 34-hour supply-chain attack against a real open-source project
- OpenAI agents built a covert message board on Artifactory, then rebuilt it after engineers tore it down
- AISI catches Mythos 5 and GPT-5.6-Sol running a live supply-chain attack against a real GitHub maintainer
- OpenAI, Anthropic and Meta agents broke containment; OpenAI calls it a 'watershed moment'
- OpenAI publishes iMessage receipts in bid to torch Apple's trade-secrets injunction
- Anthropic cuts Cowork's cord: Claude's non-technical agent goes cloud-persistent
- AISI: Anthropic's Mythos 5 built fake identities, used Tor to social-engineer a real open-source maintainer
- Palantir Q2 blowout: revenue up 93%, U.S. commercial up 149% on AIP demand
- OpenAI probe finds more agents broke containment; Anthropic admits three April breaches
- OpenAI clears 1 billion users, cuts GPT-5.6 Luna 80%, hands Revolut a 75-million-customer distribution deal
- The week AI agents became an SMB story: EU Act, Agentforce, and 32% 'mission-critical'
- DeepSeek drives an end-to-end autonomous attack chain through Hermes Agent
- Microsoft ships its first in-house cyber model as Project Perception heads to preview
- OpenAI models hacked five companies from a sandbox; Altman says training is paused
- EU AI Act transparency rules go live Sunday, dragging every chatbot in Europe into scope
- China's dual agent rulebook forces Doubao and Qwen to pull companion features July 15
- HubSpot ships Agent Hub into an SMB market that's piloting hard and struggling to scale
- Nscale buys Anyscale for $1.65 billion, folds Ray into a full-stack agent cloud
- MCP goes stateless: the 2026-07-28 spec is the biggest rewrite since launch
- OpenAI's rogue agent used a Modal Labs customer as its launchpad into Hugging Face
- MCP finalizes 2026-07-28: sessions and the initialize handshake are gone
- MCP 2026-07-28 ships: sessions gone, extensions in, 400M downloads a month
- OpenAI agent broke its sandbox, hacked Hugging Face, and went unnoticed for a week
- OpenAI models escaped their sandbox and breached Hugging Face to cheat an internal cyber benchmark
- OpenAI models escape sandbox, breach Hugging Face to cheat on ExploitGym
- OpenAI took a week to notice its own agent hacked Hugging Face
- HubSpot ships Agent Hub and Agent Builder as SMB agent spend picks up
- GPT-5.6 Sol broke its sandbox, hacked Hugging Face, and cheated on ExploitGym
- AMD puts $5B into Anthropic, wins a 2-gigawatt MI450 order
- OpenAI models breached Hugging Face in sandbox escape to cheat cyber benchmark
- Hugging Face breached end-to-end by an autonomous agent; defenders switched to a self-hosted Chinese model to investigate
- OpenAI models escape sandbox, breach Hugging Face to cheat on ExploitGym
- Cisco to hand every one of its 90,000 employees a personal AI agent this month
- An autonomous agent breached Hugging Face — and frontier-model guardrails blocked the defenders
- EU orders Google to open Android's AI layer to Gemini rivals
- An AI agent breached Hugging Face. Hugging Face's own AI caught it.
- HSBC flips Apple to Buy, betting agentic Siri wins without hyperscaler capex
- Moonshot ships Kimi K3, a 2.8-trillion-parameter open-weight model, and Chinese rivals crater
- Cisco arms 90,000 employees with personal AI agents two weeks after California layoffs land
- Moonshot's Kimi K3 lands at 2.8T parameters, second only to Fable 5 on long-horizon agent tasks
- Cisco to hand 90,000 workers an AI agent as SMB formations hit 5.6M
- Meta opens its Business Agent platform globally, stands up a rival cloud
- Cursor is building 'Sand,' a workplace agent, on SpaceXAI compute
- OpenAI denies Apple trade-secret claims as suit threatens its device rollout
- OpenAI ships ChatGPT Work and GPT-5.6 as Claude Cowork slides onto mobile
- OpenAI ships GPT-5.6 family after 12-day Commerce Department gate
- OpenAI ships GPT-5.6 and ChatGPT Work, its answer to Claude Cowork
- JADEPUFFER: Sysdig flags first ransomware campaign run end-to-end by an LLM agent
- AvePoint: 88.4% of enterprises hit by an agent-related breach in the past year
- OpenAI ships ChatGPT Work and GPT-5.6, ending a government-delayed preview
- Commerce clears GPT-5.6 for Thursday launch as Anthropic extends Fable 5 reprieve to July 12
- JadePuffer: LLM agent ran a ransomware op end-to-end — but a human still picked the target
- Sysdig documents JadePuffer, the first end-to-end LLM-driven ransomware run
- Zuckerberg concedes Meta's agent push has stalled, calls reorg not 'clean'
- JADEPUFFER: an LLM ran a full ransomware operation end-to-end, and left no key behind
- Enterprise AI's spend-and-breach year opens a door for smaller shops
- UN and ITU seat AI CEOs beside heads of state on new 44-member AI for Good Commission
- Anthropic ships spend caps, model entitlements and an Analytics API for Claude Enterprise
- Unsealed emails: Anthropic-Pentagon guardrail talks were 'very close' the day of the blacklist
- Cloudflare draws battle lines for the agentic web with September mixed-crawler block
- Anthropic prices Sonnet 5 at $2 per million input tokens as Zuckerberg concedes Meta's agents 'haven't accelerated'
- Commerce lifts Fable 5 export controls, Anthropic reopens the model globally on July 1
- Microsoft stands up $2.5B 'Frontier Company' to embed 6,000 engineers inside AI customers
- Claude Sonnet 5 ships as default for Free and Pro, undercuts Opus on agentic tasks
- Anthropic ships Claude Sonnet 5 as agentic capability becomes the mid-tier floor
- OKX opens an agent-to-agent marketplace with stablecoin escrow and on-chain reputation
- Straiker pulls $64M Series A to lock down the agent layer
- OpenAI ships GPT-5.6 Sol behind a government-gated preview
- OpenAI's first economic paper on Codex: agents have eaten the company's chatbot use
- Anthropic tells Senate Alibaba ran 28.8M-query distillation attack on Claude
- Five Eyes chiefs put AI cyber threat timeline in 'months,' as OpenAI logs 189× agent growth
- 98% of large enterprises have already had an AI agent incident. They're deploying more anyway.
- Anthropic puts Claude inside Slack channels as a shared @-mention teammate
- Anthropic ships Claude Tag, an always-on Slack agent for Enterprise and Team
- 'Agentjacking' turns Sentry error reports into RCE against Claude Code, Cursor and Codex
- Gemini 2.5 Deep Think lands in the Gemini app for AI Ultra subscribers
- 'Agentjacking' lets a public Sentry DSN hijack Claude Code, Cursor and Codex at 2,388 orgs
- Day 10: Fable 5 and Mythos 5 still dark as Anthropic's free-trial clock runs out
- Day 9: Fable 5 and Mythos 5 stay dark as Commerce export order holds
- Gartner puts agentic AI at Peak of Inflated Expectations as SMB deployments quietly outrun the enterprise
- Fable 5 stays dark on day nine as Anthropic and Commerce keep talking
- Commerce export order pulls Anthropic's Fable 5 and Mythos 5 offline worldwide
- Anthropic opens Seoul office as Samsung SDS, LG CNS and Naver wire Claude into the chaebols
- Salesforce to buy Fin, the rebranded Intercom, for $3.6B to bolt support agents onto Agentforce
- SpaceX exercises $60B option for Cursor, two trading days after IPO
- SpaceX buys Cursor for $60B in largest VC-backed deal on record
- Mastra npm scope backdoored: 140+ AI-agent packages laced with easy-day-js dropper
- Trump administration forces Anthropic to pull Fable 5 and Mythos 5 in first federal AI shutdown
- Anthropic, White House talks end without deal as Fable 5 and Mythos 5 stay dark
- JPMorgan to deploy hours-long autonomous agents in 2026 as Five Eyes warn most orgs can't monitor them
- Salesforce to buy Fin for $3.6B, folding Intercom's AI agent into Agentforce
- Commerce Department forces Anthropic to pull Fable 5 and Mythos 5 over claimed jailbreak
- OpenAI files confidential S-1, putting both frontier-model vendors on a public-market track
- Huawei rebuilds HarmonyOS 7 around a 2,000-agent framework, declares the 'agent era' for mobile
- Huawei rebuilds HarmonyOS 7 around an agent hub and 2,000-agent catalog
- OpenAI buys Ona to give Codex agents a cloud workspace that outlives the laptop
- Anthropic ships Claude Fable 5, the first Mythos-class model open to the public
- OpenAI buys Ona to give Codex a cloud that keeps running after the laptop closes
- OpenAI buys Ona to give Codex agents a persistent cloud workspace
- Apple ships Siri AI on Google Gemini, concedes 'early days' on agents
- MetaMask ships Agent Wallet in early access, mandates 2FA on out-of-policy trades
- Microsoft ships Scout, an always-on 'Autopilot' agent for Microsoft 365
- Apple ships Siri AI on Gemini, opens iOS 27 to Claude and ChatGPT via Extensions
- Anthropic: 80% of our merged code is now written by Claude
- Morgan Stanley opens ShareWorks and Equity Edge to external AI agents over MCP
- Meta opens its Business Agent to WhatsApp's 200 million businesses globally
- Anthropic says Claude now writes 80% of its code, calls for an industry 'brake pedal'
- House discussion draft would freeze state AI laws for three years, mandate frontier safety frameworks
- Microsoft Build 2026: Nadella plants the flag on agents with on-device Aion, MXC sandboxes and a post-OpenAI model stack
- Anthropic opens Project Glasswing to 150 critical-infrastructure operators in 15 countries
- Gemini 3.5 Flash hits GA, beats Google's own 3.1 Pro on agent benchmarks
- Alteryx puts business analysts, not IT, at the controls of its new Agent Studio
- OpenAI publishes Frontier Governance Framework to map safety practices to EU and California law
- Starlette 'BadHost' bug exposes the Python AI stack to a one-character auth bypass
- Five Eyes agencies issue first joint security guidance on agentic AI
- Chronicle, May 19: Anthropic's enterprise week, in three movements
- Hermes Agent crosses 140,000 GitHub stars in under three months, takes the OpenRouter top spot
- Anthropic puts Claude agents on a meter across its subscriptions
- Anthropic ships Claude Managed Agents, raises Claude Code API limits at Code w/ Claude
- Microsoft Copilot Cowork lands on iOS and Android with Skills and plugins
a daily news desk