Autonomous AI that acts, not just answers
AI agents are autonomous systems that can plan, reason, and take actions across software and the web without continuous human instruction. Unlike chatbots that answer questions, agents complete entire workflows: booking meetings, writing and executing code, processing invoices, and coordinating with other agents. The enterprise adoption curve is accelerating, with Databricks reporting a 12x increase in production agent deployments in Q2 2026.
OpenAI released three GPT-5.6 tiers on July 9: Sol, Terra, Luna priced at $5, $2.50, $1 per million tokens, undercutting Anthropic and signaling pricing war.
SpaceX's $60B Cursor acquisition powers Grok 4.5 at $2/$6 per million tokens, targeting developers with proprietary IDE training data competitors cannot replicate.
OpenAI's pre-release test method replays 1.3M conversations through candidate models, catching safety gaps red teams miss.
Claude Fable 5 launches at $10/$50 per token, half the price of prior Mythos and with visible guardrails that trigger fallback to Opus 4.8 for high-risk requests.
Databricks acquires Panther Labs, targeting the security market with AI agents that automate threat detection and SOC response end to end.
Sakana AI launches Marlin, an 8-hour autonomous research agent that compresses weeks of C-suite strategy analysis into a single unattended session.
Meta's AI Mode uses public Facebook posts to answer search queries, putting 3.3 billion users' content to work in the AI platform race.
Salesforce acquires Fin AI for $3.6 billion, betting enterprise customer service shifts from human agents to autonomous AI as CRM market faces structural disruption.
GPTZero found only 5 of 45 citations in KPMG's AI report were real. UBS, NHS, and Swiss Railways denied the claims. KPMG pulled the document.
Anthropic launched Claude Design after inviting Figma and Canva to co-announce it, then dropped them, in a move that challenges both companies directly.
Satya Nadella argues proprietary AI feedback loops build durable advantage, while enterprises that only rent frontier model access will fall behind.
Anthropic leads US AI subscriptions at 41%, topping OpenAI for a second month as Claude Code now accounts for 4% of all GitHub commits in enterprise teams.
Google OKF v0.1 lets any AI agent read enterprise knowledge bases in plain markdown with no vendor SDKs required, aiming to break AI platform lock-in.
Anthropic's June 15 billing change moves Claude Agent SDK to full API rates, ending subsidies that let developers extract $175 in API value per $20.
Anthropic's new /fork command lets Claude Code branch any session into parallel variants, cutting the cost of wrong turns in long agent workflows.
Huawei HarmonyOS 7 launches Agent Framework 2.0 with 90 percent task completion and 2,000 integrated AI agents in a major push for ecosystem independence.
Meta hit billions in internal AI spend after staff burned 73.7 trillion tokens in 30 days, and now it's capping access to Claude across teams.
Moonshot AI's Kimi K2.7 Code delivers 30% fewer reasoning tokens and 21.8% higher coding benchmark scores, open-sourced under MIT license on June 12.
Kimi Work by Moonshot AI brings 300 parallel sub-agents to macOS and Windows, running K2.6 to automate complex research and office tasks without cloud.
Anthropic made DXC a Global Premier Partner on June 11 to certify tens of thousands of engineers for banks, airlines, and insurers in 90 days.
Visa embedded its payment network inside ChatGPT, letting AI agents complete purchases at any Visa merchant with user-controlled spending limits.
Microsoft Agent 365 brings AI agent governance to all 276,000 KPMG staff, marking the Big Four's first full enterprise-wide AI agent deployment.
Decart's Oasis 3 generates photorealistic multi-camera driving environments via API at $0.02 per second, backed by $300M from Nvidia and Toyota.
Mastercard's AP4M gives AI agents their own blockchain credentials and payment rails, letting machines spend across Stripe, Coinbase and 29 more platforms.
Cognition's Devin AI engineer surpassed 492M in ARR and raised 1B at a 26B valuation, with Goldman Sachs reporting 3x to 4x productivity gains.
Anthropic's Claude Platform now runs agents on cron schedules with vault-stored credentials, making autonomous AI workers available to any enterprise team.
Huawei Cloud at INSPIRE 2026 launched AgentArts and 200 EFLOPS Agentic Infra, choosing enterprise productivity over token pricing cuts as rivals battle on cost.
Anthropic's Fable 5 tops SWE-Bench Pro at 80.3%, beating GPT-5.5 by 22 points, opening the Mythos-class tier at $10 per million tokens.
OpenAI's o1 model scored 67% accuracy on 76 real ER triage cases versus 55% for physicians in a Harvard study published in Science in April 2026.
Google Dreambeans uses Gmail, Calendar, Photos, and YouTube data to generate 10 to 14 personalized AI stories daily. Privacy is the unavoidable question.