Claude news June 2026
114 updates from Claude and Anthropic in June 2026, from product launches to research papers, each with a short satori. note on what it means for a European business.
We've received notice that the Department of Commerce has lifted export controls on Claude Fable 5
and Mythos 5. We'll begin restoring access tomorrow, and will share an update soon.
The export controls are lifted and Fable 5 and Mythos 5 are coming back. European companies should read it as a sign that the dialogue between the US government and Anthropic works, while the risk remains. Build with that in mind.
We've received notice that the Department of Commerce has lifted export controls on Claude Fable 5 a
nd Mythos 5. We'll begin restoring access tomorrow, and will share an update soon.
The news that saved the third quarter for many teams. The end of roughly four weeks in which many of our clients ran a workaround through Bedrock.
Early access partners found Sonnet 5 finishes complex tasks where previous Sonnets stopped short, ch
ecks its own output without being asked, and does all its agentic work at an attractive price point.
Sonnet 5 double-checks itself without being told to. That is what separates a prototype from a product. If the agent verifies its own work, teams avoid building a whole QA round around it.
Sonnet 5 is now the default on Free and Pro, and available to Max, Team, and Enterprise users.
It's live across all Claude apps and the Claude Platform today, with introductory pricing through August 31.
Sonnet 5 is the default on Free and Pro. That means everyone you talk to is already using it without knowing. Recalibrate prompts and output expectations before August is over.
Introducing Claude Sonnet 5, our most agentic Sonnet yet.
It makes plans, uses tools like browsers and terminals, and runs autonomously at a level that just a few months ago required larger and more expensive models.
Sonnet 5 lives up to the hype, it is agentic, makes plans and runs autonomously. For teams already building workflows on Sonnet the upgrade is a no-brainer. Run your pipelines through it this week.
Introducing Claude Sonnet 5
Sonnet 5 is Anthropic's most agentic Sonnet so far. The model plans, uses a browser and a terminal, and runs autonomously at levels that used to require larger models.
Sonnet 5 runs autonomously at what used to be Opus level, at a Sonnet price. That is where small and mid-sized businesses start to make real agents pay. The introductory price until 31 August is a good window to test in production.
Introducing Claude Science, a new app designed with every stage of research in mind.
Artifacts traced to their code, environments managed on demand, and 60+ optional scientific databases that you can connect. Available now in beta.
Claude Science is a tool for researchers. For universities and R&D departments it is a chance to escape spreadsheet nightmares and work reproducibly. Beta now, pay later.
Redeploying Fable 5
Fable 5 is relaunched globally after US export controls were lifted, now with new classifiers that block more advanced cybersecurity tasks.
Fable 5 is back, now with firmer guardrails. For European customers it shows that Anthropic can handle political pressure without stopping product development. Continuity holds for now, still build with a plan B.
Claude Science, an AI workbench for scientists, is now available
Anthropic launches Claude Science, an AI tool for researchers that brings scientific databases, data analysis and compute resources into a single environment.
Claude Science is built for researchers, not for everyone. The idea of 60 databases in one flow is something corporate R&D departments should look at too, not just academia.
Inference runs on Azure infrastructure, operated by Anthropic.
Prompt caching and extended thinking are supported today, with more capabilities on the way.
Prompt caching and extended thinking in Azure from day one. Technically faster and cheaper. Practically, one invoice less to explain to the CFO.
Claude in Microsoft Foundry is now generally available, hosted on Azure.
Azure customers get Claude Opus 4.8 and Claude Haiku 4.5, with Azure authentication, billing, and commitment retirement.
Claude in Azure is generally available. For companies on a Microsoft stack, procurement suddenly gets easier because Claude sits inside an existing cloud contract. Anthropic is really cutting the friction with enterprise IT.
Claude in Microsoft Foundry is now generally available, hosted on Azure and operated by Anthropic.
Azure customers can build production applications and agents with Claude Opus 4.8 and Claude Haiku 4.5, all within their Azure environment. It works with the
Claude in Microsoft Foundry is generally available on Azure. For companies already living in Azure it means Claude Opus 4.8 and Haiku 4.5 without touching the billing setup. A reasonable shortcut for Azure-heavy customers.
Since June 12, we've been working closely with the US government to restore access to Claude Mythos
5 and Fable 5. Today, the government notified us that Mythos 5, our strongest cybersecurity model, can be redeployed to a set of US organizations
Mythos 5 is back, but only for selected US organisations defending critical infrastructure. In Europe we are still working with Opus and Sonnet. That works.
Anthropic Economic Index report: Cadences
The Anthropic Economic Index reports how Claude usage follows daily patterns, with personal use peaking at weekends and work-related use dominating weekdays.
Anthropic measures Claude usage hour by hour. The pattern is clear, work on weekdays and personal use at weekends. A good signal for anyone planning AI rollouts around office hours.
Claude now works with public health data through two new additions to the Claude connector directory
. - Near-real-time US health data and insights across a variety of conditions including respiratory illness, immunizations, chronic disease, injury and overdoses, and more. Hosted
Two new health data connectors in Claude. For regional health authorities it shows that the right API connection can make Claude an analysis partner, not just a chatbot.
AI's impact on the economy will ultimately show up in aggregate data like employment and productivit
y. But it will first be visible where AI is doing the most work. By tracking how usage shifts, hour by hour, surface by surface,
Measuring AI effects where the work actually happens is smarter than waiting for GDP figures. We use the same logic when we measure project value with clients.
Nearly half of respondents expect their work responsibilities to significantly change in the next 12
months. Fewer than 10% think they'll lose their own job within a year, but far more worry for coworkers: over 1/3 put the odds of
Half expect big changes within 12 months, and a third worry about their colleagues' jobs. A genuine dilemma. We talk about it with clients more than ever.
To keep pace with AI progress, we're advancing how we study Claude's economic impact.
Hourly sampling and survey data show us how the cadences of life shape usage, what people produce with Claude, and how perceptions of AI's impact
Anthropic measures Claude usage hour by hour. Good methodology, and we would like the same transparency from other suppliers. We follow the Economic Index closely.
We're joining RAISE US as a founding partner.
RAISE US is a nonprofit coalition working to strengthen the American workforce through employer-led action, AI-enabled training, and policy innovation to support the transition to
Anthropic takes the US labour market seriously. A European equivalent is missing, and someone should start one this autumn.
We are joining RAISE US as a founding partner.
RAISE US is a nonprofit coalition working to strengthen the American workforce through employer-led action, AI-enabled training, and policy innovation to support the transition to
Anthropic joins RAISE US as a founding partner on AI and the labour market. The focus is American, yet this kind of partnership is exactly what is missing between European governments, industry and AI suppliers.
We're joining @raiseus_ai as a founding partner.
RAISE US is a nonprofit coalition working to strengthen the American workforce through employer-led action, AI-enabled training, and policy innovation to support the transition to
Anthropic joins RAISE US as a founding partner to reskill the American workforce. Europe needs something similar, and it is still unclear who will drive it.
RAISE US is a national nonprofit building America's people strategy for the AI economy, the workforc
e infrastructure the country needs so workers can train, transition, and thrive.
RAISE US lays out its reskilling plan for the AI era in concrete terms, with a lot of detail. We would like the same clarity in national AI strategies across Europe.
RAISE US is a national nonprofit building America's people strategy for the AI economy.
The workforce infrastructure the country needs so workers can train, transition, and thrive. We are co-chaired by Gina Raimondo, the 40th U.S. Secretary of Commerce
RAISE US is building America's AI labour market strategy with Gina Raimondo and Eric Holcomb at the top. Europe has nothing comparable yet, and the open question is who should take the initiative here.
In a single-player AI experience, you connect your own accounts and the agent acts on your behalf.
Claude Tag is multiplayer, so of the multiple people in a conversation, whose identity should it use? Turns out: its own. On Slack, Claude Tag
A smart answer to multiplayer AI, Claude has its own identity in channels and borrows yours in DMs. It handles exactly the dilemma we have discussed with several clients.
Leaders from pharma, biotech, and research are getting together to talk about how they're building w
ith Claude in the lab. Watch the livestream on June 30 at 10am PT: https://lnkd.in/gyi7PuJ5
Pharma and biotech gather around Claude in the lab. Europe has strong life sciences companies that are moving slower than their US peers, and it is an area where we want to see more movement.
Turn ambient behavior on, and Claude takes initiative.
It follows up on threads that have gone quiet and flags what's relevant from across its channels and tools.
Claude proactively following up forgotten threads is the dream for anyone leading a team. The risk is noise if you do not tune it. Start narrow.
Claude Tag is available in beta today on Slack for Claude Enterprise and Team plans.
We plan to expand it more widely in the future. Read more:
Only Enterprise and Team in beta. For small businesses that means waiting a little longer. The trend is crystal clear though, AI is moving to where the conversation happens.
Introducing Claude Tag, a new way for teams to work with Claude.
In Slack, Claude joins as a team member with access to the channels and tools you choose. Tag Claude in and delegate tasks to it
The Slack version of Claude lands today. We have been waiting for this, because a lot of our clients' daily work actually happens in Slack, not in a chatbot tab.
Introducing Claude Tag, a new way for teams to work with Claude.
In Slack, Claude joins as a team member with access to the channels and tools you choose. Tag Claude in and delegate tasks to it
Anthropic backs Claude Tag. Multiplayer AI in Slack is a serious change of game for teams, and we already have a queue of clients who want to test it.
Introducing Claude Tag
Claude Tag is Anthropic's new Slack integration that lets teams delegate tasks to Claude, which learns from channel context and works autonomously.
Claude moves into Slack as a real team member. For many companies it is the first time AI actually sits where the work happens, not in a separate tab.
As your session keeps working, the artifact refreshes, so everyone you've shared it with is always l
ooking at the latest version.
Live dashboards that update as Claude keeps working. They replace the pseudo status reports consultants like to email on Thursday evenings.
Artifacts draw on the full context of your session: codebase, plugins, skills, connected tools.
They're private until you share them, and sharing stays inside your organization. Read more:
Artifacts inherit the full session context and stay private until you share them. A good default, since many people we talk to have been nervous about leaks.
New in Claude Code: Artifacts.
Interactive pages built from your session, like a PR walkthrough or a living project dashboard, shared with your team at a private link. Available in
Interactive pages from a Claude Code session, shareable with the team. Promising for consulting work where you want to show the client what the agent actually did.
Project Fetch: Phase two
Claude Opus 4.7 performed roughly 20 times faster than human teams on robotics tasks, showing progress in the physical capabilities of AI.
The robot is now steered by a good prompt. It means warehousing, manufacturing and field work are no longer a magic safe zone, even if industrial jobs in Europe are secure for a while yet.
Watch the robodogs in action in our first Project Fetch experiment:
Robodog demos are fun, but the real value is the insight that frontier models improve faster than the hardware. That is where the focus should be.
New Frontier Red Team blog: Phase 2 of Project Fetch, where we test how well Claude can program a ro
bodog. Opus 4.7, on its own, was ~20x faster than last year's best human team aided by Opus 4.1. (The robodog, alas, still failed to
Opus 4.7 is 20 times faster than last year's Opus 4.1, and the robot dog still missed the ball. A good reminder that model benchmarks do not always lead to real results.
Anthropic opens Seoul office and announces new partnerships across the Korean AI ecosystem
Anthropic opens an office in Seoul and starts partnerships with the Korean government and industry on AI safety.
The expansion in Asia continues. For European buyers in regulated sectors, a local partner still does most of the groundwork of fitting Claude to GDPR and the EU AI Act.
Claude Design and Claude Code work together in both directions (rolling out today).
Hand a design off to build, or start in Claude Code and sync design projects from your terminal. Export to PDF and PowerPoint, or send
Two-way sync between design and code. It is a small step, yet it removes a lot of the friction between handoff and build. We are testing it internally.
Claude Design is in beta on all paid plans, on web and desktop.
Give it a try:
Design opens up broadly, including on Pro. An inclusive rollout, which is unusual, and it means solo designers can start testing today.
New in Claude Design: it stays on brand with your design system across projects, lets you edit direc
tly on the canvas, syncs with Claude Code, and connects to more of the tools you already use.
Design that follows the client's design system and syncs with Claude Code. For frontend builders the step between mockup and production gets a little shorter again.
These and other measures will allow us to track consequential shifts in the nature of work as they h
appen, we'll incorporate some of them into the Anthropic Economic Index going forward. Read the full report:
Anthropic measures the value of AI work continuously and publishes it openly. That transparency is worth more than any marketing budget.
Domain experts, as judged by the questions they ask and vocabulary they use about a subject, are more
likely to see success. But the gap between intermediate and expert users is quite modest, suggesting that proficiency in a domain is sufficient to code…
Domain knowledge beats coding skill. This confirms why we are comfortable putting Claude in the hands of the CFO or the head of sales, not just developers.
Domain experts as judged by the questions they ask and vocabulary they use about a subject are more
likely to see success. But the gap between intermediate and expert users is quite modest, suggesting that proficiency in a domain is sufficient to code
Domain experts do better, but the gap to the average is small. In plain terms, you do not need to be senior to get Claude to deliver. Good news for a broad rollout.
These and other measures will allow us to track consequential shifts in the nature of work as they h
appen, we'll incorporate some of them into the Anthropic Economic Index going forward. Read the full report:
Anthropic promises more measurement over time. Good for those of us advising on AI rollouts, because data-driven arguments beat emotional ones in every industry.
The average task in Claude Code has grown more valuable.
We compared the type of work done in each session to what that same task would cost on a freelance marketplace. From October to April,…
27 percent growth in value in six months. That is not the hype curve, it is substance, and it now shows up in our clients' invoices.
We compared Claude Code success rates between occupations.
On our toughest measure of success, requiring verifiable evidence that a goal was completed, like committed code, every field was within 7 percentage points of software engineering.
Seven percentage points between a software engineer and someone in law or finance. It means the whole office can become builders, not just the IT department.
Our latest economic research introduces a framework for tracking Claude Code as it scales.
Who is using Claude Code, and what are they using it for? How is the value of tasks changing? And how much does domain expertise…
A good framework to lean on when we pitch to finance directors. Anthropic does the work for us with hard data rather than vibes.
Our latest economic research introduces a framework for tracking Claude Code as it scales.
Who is using Claude Code, and what are they using it for? How is the value of tasks changing? And how much does domain expertise
Who uses Claude Code and for what. Interesting to see how junior developers perform against senior ones, and we use the data when we coach client teams.
Code with Claude just wrapped across SF, London, and Tokyo.
We asked builders at every stop: what's the one thing you're taking back to your own work?
Code with Claude ran in San Francisco, London and Tokyo. London is the one European stop so far, and until more cities follow we run our own versions with clients.
Agentic coding and persistent returns to expertise
Domain expertise, not coding skill, decides success with AI coding tools. Every occupational group succeeds almost as well as software engineers.
This is gold for businesses. People who know the industry can now build, even without being developers. The heads of finance, sales and support can suddenly ship code.
The US government, citing national security authorities, has issued an export control directive to s
uspend all access to Fable 5 and Mythos 5 by any foreign national, whether inside or outside the United States. The net effect of this…
The same export control decision, now shared widely on LinkedIn. Worth noting that Anthropic communicates openly when models disappear. That is why we are comfortable relying on them in client builds.
The US government, citing national security authorities, has issued an export control directive to s
uspend all access to Fable 5 and Mythos 5 by any foreign national, whether inside or outside the United States, including foreign national Anthropic employees.
A US export control switches off Fable 5 and Mythos 5 for foreign nationals. Put this next to the GDPR question when you choose an AI supplier, because political risk is as concrete as data storage.
Statement on the US government directive to suspend access to Fable 5 and Mythos 5
The US government has ordered Anthropic to shut off access to Fable 5 and Mythos 5 on security grounds, despite the company's objections.
This is why every Claude setup we build for European clients comes with a plan B. When US export policy switches models off overnight, you do not want to be left without an alternative model ready to take over.
Results from the first Anthropic Public Record
Anthropic's survey of 52,000 Americans finds that half hope AI will cure diseases, 64% fear job losses and only 15% trust AI companies.
Only 15 percent trust AI companies. That is not a numbers problem, it is a positioning problem, and it is why companies that dare to publish their AI policies win trust faster.
TCS and Anthropic partner to bring Claude to regulated industries
Anthropic and TCS start a partnership to bring Claude to regulated industries globally, with a focus on finance and healthcare.
TCS and Anthropic move into finance and healthcare together. For European companies in the same sectors it means Claude now has a partner that already knows the compliance nightmares.
We are conducting a new survey series, Anthropic Public Record, to understand how the public thinks
and feels about AI, and presenting a snapshot of the results from the first wave, fielded in November and December of 2025 with nearly 52,000…
Anthropic measures public feeling about AI with 52,000 responses. We run the same exercise with small and mid-sized clients every week, and the picture is the same, hope mixed with a lot of uncertainty.
What are you building?
A small question with big signal value. Claude is building a community, not just a product, and it shows in how quickly new features appear.
Mythos is going crazy with the glass liquid UI with just css x.
com/claudeai/statu…
A glass liquid UI with nothing but CSS. For an online retailer it means prototypes in minutes instead of design sprints.
Claude Fable 5 has been out for a couple of days.
Some projects people have already built with it:
Fable 5 has only been out for a few days and the community is already delivering. That speed is what makes Claude hard to beat right now.
A soccer game, a neo-gothic city, a plane model, and other projects built with Claude Fable 5.
Football games, neo-gothic cities and aircraft models. Demos are fun but uninteresting until someone shows the same trick on an invoice pipeline or a quote template. That is where we should look.
We sat down with HubSpot's product and marketing leaders to ask what's different since their team…
started using Claude Cowork. What they told us: https://lnkd.in/e4TPUN7U
HubSpot's product and marketing leaders explain what has actually changed with Claude Cowork. Worth reading before the leadership team asks you for ROI evidence from real companies.
DXC will integrate Claude into the systems banks, airlines, and other regulated industries rely on
Anthropic and DXC Technology start a global collaboration to integrate Claude into critical systems for banks, airlines and other regulated industries.
Claude moves into the core systems of banks and airlines through DXC. For regulated European sectors this is confirmation that Claude is being built for compliance, not just chat.
We're launching Claude Corps, a national fellowship program matching people early in their career…
with US nonprofits. We'll teach 1,000 people to use Claude, and pay them to use AI to advance their hosts' missions.
Anthropic is creating a pipeline of young AI-trained people who will work hands-on with real organisations. A smart move, the next generation of decision makers will have Claude in their muscle memory.
We're launching Claude Corps, a fellowship program for people early in their careers who are pass…
onate about extending the benefits of AI to communities across America. Alongside CodePath and Social Finance, we'll teach 1,000 people how to use Claude, match…
Claude Corps places 1,000 young people in non-profits to roll out AI for real. Strategic grants like this let Anthropic build its brand and its talent pipeline at the same time.
Introducing Claude Corps
Anthropic launches Claude Corps, a nationwide fellowship programme that places 1,000 young professionals in non-profit organisations to use AI tools and build skills.
A fellowship for 1,000 young people working pro bono with Claude at non-profits. An interesting signal that Anthropic wants to build an AI-fluent workforce rather than just sell licences.
By themselves, these projects will not be sufficient to meet the challenge of advanced AI.
But they’re a signal of our intent. Over the coming months and years, we will expand our work on these fronts much further.
A founder admitting that the initiatives are not enough is unusual and useful. Companies claiming that one AI pilot will fix everything could learn from that humility.
And tomorrow, we’ll launch a $150 million national fellowship program designed to help people early
in their careers extend the benefits of AI to communities across America.
A 150 million dollar fellowship is PR, of course, but the structure says something, money aimed at people early in their careers. A European programme half as ambitious would already be a big step.
AI is advancing at a pace our policymaking institutions were never built for, and the gap between the
two is becoming the central challenge of the technology. In his latest essay, our CEO Dario Amodei lays out how to close it. We’re launching
Dario points out the policy gap, which is worth an afternoon for any leadership team. Not to copy the US, but to avoid missing the window that EU AI Act implementation gives us.
Michael T.
fell in love with coding at 12. The company he co-founded, Cursor, went from 15 people to 700 in two years. Today, over 60% of
The Cursor story on LinkedIn makes the same point as on X in a different channel. Reach is everything, and the Claude team is deliberately building its distribution muscle.
Michael Truell ( @mntruell ) fell in love with coding at 12.
The company he co-founded, @cursor_ai , went from 15 people to 700 in two years. Today, over 60% of the Fortune 500 build with its
Cursor went from 15 to 700 people in two years and is used by 60% of the Fortune 500. That is the speed European SaaS companies now compete against, not the local rival down the street.
From The Problem Solvers, our series featuring founders taking on hard problems with Claude: https:/
/ claude.com/problem-solvers
The Problem Solvers series is free research material for CEOs wondering how to build an AI-heavy company. More value there than in a consultant's 80-page report.
New from Code with Claude Tokyo: scheduled deployments and environment variables in vaults are now i
n public beta in Claude Managed Agents, and dynamic workflows in Claude Code are now generally available. - A scheduled deployment runs your agent on
Scheduled agents and vault keys are operations features that CTOs should stop waving away as boring. They are the difference between a proof of concept and production.
Scheduled deployments and environment variables in vaults are available today on the Claude Platform
. Read more:
Scheduled deployments and vaults for environment variables are now live on the Claude Platform. Suddenly this starts to look like a real production environment, not just a sandbox.
Dynamic workflows in Claude Code are now generally available
. For complex tasks like codebase-wide bug hunts, Claude writes its own orchestration and runs subagents in parallel, verifying the work before it reaches you.…
Dynamic workflows in Claude Code are generally available. Claude orchestrates its own subagents and double-checks the work before you see it. The threshold for chasing bugs across a whole codebase just dropped a lot.
New from Code with Claude Tokyo: scheduled deployments and environment variables in vaults are in pu
blic beta in Claude Managed Agents, and dynamic workflows in Claude Code are generally available. Agents now run on a schedule, use your tools securely,
Scheduled agents and vault-based keys are the boring things that actually make AI operable. Companies that have not built this yet should start here.
An Anthropic sales leader runs his entire 4,000-account book with Claude Cowork, including morning c
all briefs, Friday forecasts, and quarterly scoring across every account. How he set it up: https://lnkd.in/dq-ceCDb
4,000 accounts on one salesperson with Cowork shows that scaling operations through AI is no longer a pitch. B2B sales teams that do not work this way are already behind.
Introducing Claude Fable 5: a Mythos-class model that we've made safe for general use.
Its capabilities exceed those of any model we've ever made generally available. Fable 5 is state-of-the-art on nearly all tested benchmarks, with exceptional performance in
Fable 5 on LinkedIn with a research focus. This is Anthropic's enterprise flank, research-driven and safety-focused, less hype and more substance. That is how we talk to CIOs.
Claude Fable 5 is available everywhere today.
Claude Mythos 5 is restricted to Glasswing partners until we expand our trusted access program.
Fable 5 is available everywhere now, Mythos 5 only through Glasswing. For those of us rolling out Claude for real, Fable 5 is more than enough, and Mythos is overkill for 99 percent of client work.
Soon, we intend to expand access to Mythos 5 through a broader trusted access program, both for defe
nsive cybersecurity work and biomedical research.
Mythos 5 opens up for defence security and biomedicine. It is not a model for small businesses yet, but the signal is clear that heavy industries will soon get their own channels.
Introducing Claude Fable 5: a Mythos-class model that we’ve made safe for general use.
Its capabilities exceed those of any model we’ve ever made generally available.
Fable 5 is more than a new SKU, it carries a signal. Anthropic runs a two-track strategy between most capable and safe, and that is worth keeping in mind when you design your own AI stack.
Introducing Claude Fable 5: a Mythos-class model that we’ve made safe for general use.
Its capabilities exceed those of any model we’ve ever made generally available.
Anthropic's official account reposts its own product launch. Not strange, but the signal is that Fable 5 is the story of the week, not a side note.
Claude Fable 5 and Claude Mythos 5
Anthropic launches Claude Fable 5, a highly capable AI model with safety limits, and Claude Mythos 5 for cybersecurity professionals.
Fable 5 is the new baseline and Mythos 5 goes to the security people. For companies it means the flagship models now come in two flavours, so choose based on your risk profile.
Final stop: Tokyo.
Register to hear directly from the teams behind Claude: http:// claude.com/code-with-claude/tokyo …
Anthropic is touring the world with the final stop in Tokyo. It signals that the Asian market is a strong priority right now. Worth keeping an eye on.
When coding stops being the slow part, some long-held engineering norms are worth a second look.
Here are three defaults the Claude Code team is rethinking.
When code is no longer the bottleneck, old development habits go out of date. Worth reading for CTOs rebuilding their development processes right now.
New Science Blog: Why has AI advanced faster in coding than in biology?
To agents, bio databases are like cities built before cars, maddening to drive in because they're designed for different traffic. How do we build infrastructure
Anthropic says it plainly, AI progress races ahead in code but struggles in biology. It is a database design problem rather than a model problem, and it applies to your internal systems too.
Here's everything new in Claude from last month: - Claude Opus 4.
8 - Effort control in the model selector - Claude for Small Business - More usage in Claude Code (through July 13), Cowork (through July
Last month's Claude release notes in one place, Opus 4.8, effort control and dynamic workflows. Print it and tick off what your team has already tried at Friday's AI coffee.
Paving the way for agents in biology
AI agents struggle to retrieve virus sequences from biological databases. A deterministic tool, gget virus, raises accuracy to almost 100 percent.
The lesson is not mainly about biology. If your internal systems are built for humans, agents will stumble whatever the industry. Build tools the AI can actually get a grip on.
Measuring LLMs' impact on N-day exploits
Modern AI models can quickly turn publicly released security patches into working exploits, which dramatically shrinks the time window for defenders.
The time between patch and exploit is shrinking to hours. IT managers who still work with a monthly patch window need to rethink, preferably before the next monthly meeting.
We've doubled usage limits in Claude Cowork for the next month.
Delegate bigger, more complex tasks to Claude.
Doubled limits in Cowork for a month. A good moment to test how far you can delegate, and a free way to learn what Claude can do without hands-on help.
Live now on all paid plans through July 5.
Download the Claude desktop app to give Cowork a try: http:// claude.com/cowork
Cowork is live on all paid plans until 5 July. If you have Pro licences and have not tested it yet, do it now while the limits are extra generous.
We've doubled usage limits in Claude Cowork for the next month.
You can now delegate bigger, more complex tasks to Claude. A few things you can hand off: - Account research across dozens of companies -…
Doubled limits in Cowork for a month. Accounts, campaign reports, financial models. Those are the repetitive jobs we would actually send off in the morning and check at lunch.
Making Claude a chemist
Claude can now analyse NMR spectra as well as chemistry software and suggest molecular structures from spectral data.
Opus 4.7 matches dedicated NMR software on chemistry. That is what we mean when we say niche tools age quickly, the generalist is catching up from the side.
New Anthropic Science Blog: Making Claude a chemist.
To manipulate a molecule, chemists first need to understand its structure. Their main tool is NMR spectroscopy. We found Opus 4.7 matches, and on some tasks…
Opus 4.7 matches dedicated NMR software on chemistry tasks. This is what happens when the generalist model gets good enough to eat the niche tools' lunch.
Correction: Claude Opus 4's ~3x average speedup dates to May 2025, not May 2024.
This evaluation has only existed since September 2024, but we backtested it on earlier models: those from May 2024 showed no speedup whatsoever.
A correction on the speed-up data, which applies to May 2025 and not 2024. Good that Anthropic publishes corrections openly, because details like that build trust over time.
Strategy games, cities created with math, and other things people built with Claude Opus 4.8 this we
ek.
Strategy games and cities built with maths. Projects like these show why Opus 4.8 is not only about productivity, it is also about making it more fun to create things.
AI research is a series of next-step decisions.
We looked at sessions where a human researcher took a wrong turn, showed Claude the session up to that point, and asked it what to…
Claude improves human researchers' decisions 64% of the time. That is not automation, it is a sparring partner that sees what you miss. Valuable in far more industries than research.
None of this guarantees recursive self-improvement is on the horizon.
It's not yet clear that Claude is capable of research judgment, of choosing the right problems to work on. But if these trends continue, AI systems…
Anthropic is careful and does not say recursive self-improvement is around the corner, even if the trend points that way. Read the whole thread before drawing conclusions from a headline.
Each time we release a model, we run the same test: give it code that trains a small AI model, ask t
he new model to speed it up. It takes a skilled human 4-8 hours to reach 4x faster. In May 2024, Claude Opus 4 averaged…
A human needs 4 to 8 hours to speed up training code 4 times. Mythos Preview manages 52 times. That is no longer a difference of degree, it is a different sport.
Our internal data shows Claude is accelerating AI development, a possible path to recursive self-impr
ovement, or AI autonomously building a more capable successor. It's happening faster than we thought, and the implications deserve greater attention.
Anthropic says Claude is accelerating AI development. That is not science fiction, it is internal measurement. The question is how quickly we want to see the same thing inside our own companies.
From The Problem Solvers, our series featuring founders taking on hard problems with Claude: https:
// claude.com/problem-solvers
Anthropic's series with founders solving hard problems with Claude. Worth a look for small businesses wondering where to start, because examples are worth more than frameworks.
Anton Osika ( @antonosika ) is the co-founder and CEO of @lovable , where anyone can build software
through conversation. His working thesis: the most underrated moat in AI is trust, and earning it takes craft, care, and obsession.
Anton Osika at Lovable says trust is the underrated moat in AI. He is right, it is what separates the suppliers that survive from the ones that die in the hype cycle.
How well do the security community's techniques hold up against AI-enabled cyberattacks?
We examined 832 malicious accounts and mapped their activity onto a longstanding database of tactics and techniques used by threat actors. Here's what we learned:
832 malicious accounts mapped against established threat techniques. That is how you build real threat intelligence for the AI era, not with vague press releases about misuse.
Mapping AI-enabled cyber threats: Insights from the LLM ATT&CK Navigator
Anthropic analysed 832 accounts that misused Claude for cyberattacks, and the risk from AI-enabled threats rose from 33 percent to 56 percent over the year.
AI security is no longer an academic worry, it is a leadership question. Worth raising at the next management meeting, together with who actually owns the AI policy in your organisation.
Ben kept getting distracted during long Claude Code sessions, so he built a miniature Clawd to get h
is attention. When Claude Code needs input, a solenoid inside the aluminum body pushes it upward. In other words, it jumps up and down.
A miniature Clawd that physically jumps when Claude Code needs input. Unnecessary, yes. Charming, completely. More unnecessary things like this, please.
We spoke with @BenJames_____ about building things, creativity, and why silly ideas matter.
Ben James talks about why silly ideas matter. It is not sentimental, it is the only way to find products that actually touch people.
Introducing the Services Track and Partner Hub of the Claude Partner Network
Anthropic introduces the Services Track and Partner Hub for the Claude Partner Network, giving partners transparent criteria for building and showing their Claude expertise.
Anthropic is building a partner programme with clear criteria. That helps buyers, because it becomes easier to see which firms have real delivery experience with Claude.
What we learned mapping a year’s worth of AI-enabled cyber threats
A year of data on AI-driven cyberattacks, mapped against MITRE ATT&CK. Put the report on your CISO's desk before the next board meeting, it is free ammunition for talking about risk properly.
This Executive Order is an important step in strengthening America's leadership in AI.
We look forward to collaborating with the White House to support its implementation.
Anthropic welcomes the new executive order on AI. It is realpolitik, but the signal is that the big players work with the state rather than against it. European companies should note the tone.
Enterprise orgs can now get credits for every first-time Claude Code and Claude Cowork user, good to
use for 90 days. Terms apply: https://lnkd.in/ek4swJkK
Enterprise credits for new Claude Code and Cowork users, with a 90-day window. A smart play that lowers the threshold for letting the whole team test before the budget talk starts.
Interpreting law is one of the oldest jobs in the world.
@MaxJunestrand , co-founder and CEO of @WeAreLegora , is bringing it into its next era with Claude. His bet: every new model release raises the…
Max Junestrand is building the boats for the legal industry. The same principle applies everywhere now, every new model release raises the sea level for everyone.
From The Problem Solvers, our series featuring founders taking on hard problems with Claude: https:
// claude.com/problem-solvers
Anthropic's series with founders solving hard problems with Claude. A good format for decision makers who want to see cases before budget, not the other way round.
We're expanding Project Glasswing.
We've extended access to Claude Mythos Preview to approximately 150 additional organizations, based in more than fifteen countries. Read more about this expansion and our…
Project Glasswing expands to 150 organisations in 15 countries. We hope European security teams are on the list, because this is the kind of programme you want to join early.
Expanding Project Glasswing
Anthropic expands Project Glasswing to around 150 new organisations worldwide to secure critical software with Claude.
150 organisations get help securing their critical code with Claude. This is the direction, AI that actually hunts bugs before they end up on CVE lists.
Anthropic confidentially submits draft S-1 to the SEC
Anthropic has submitted a draft S-1 form to the SEC for a possible stock market listing.
Anthropic is preparing for a listing. For European buyers it means more maturity and more enterprise focus. To be honest, it is also a good moment to think about vendor lock-in once more.
We use analytics cookies to see which pages people read, and marketing cookies stay off unless you tick them under Details.