Future of work

BlackSuit infrastructure dismantled in global raid

US law enforcement, alongside nine other nations, dismantled the BlackSuit ransomware gang’s infrastructure, replacing its leak site with a takedown notice after a coordinated operation. The group, formerly known as Royal, had amassed over $370 million in ransoms since 2022.

More than 450 victims were targeted across critical infrastructure sectors, with ransom demands soaring up to $60 million. Dallas suffered severe disruption in a notable attack, affecting emergency services and courts.

German authorities seized key infrastructure, securing data that is now under analysis to identify further collaborators. The operation also included confiscating servers, domains and digital assets used for extortion and money laundering.

New research indicates that members of BlackSuit may already be shifting to a new ransomware operation called Chaos. US agencies seized $2.4 million in cryptocurrency linked to a Chaos affiliate, marking a significant blow to evolving cybercrime efforts.

Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot!

Musk ends in-house ‘Dojo’ AI chip programme, shifts focus to external partners

Tesla has reportedly shut down its Dojo supercomputer project following multiple high-profile departures, including that of project head Peter Bannon. CEO Elon Musk ended the AI chip programme, reassigning the remaining staff to other data centre projects.

Dojo aimed to process vehicle data for autonomous driving and reduce Tesla’s reliance on Nvidia and AMD. The project faced delays, with leaders such as Jim Keller, Ganesh Venkataramanan, and Bannon departing before its closure.

About 20 former Dojo employees have joined DensityAI, a stealth startup founded by ex-Tesla staff, which is expected to work on AI chips for robots and data centres. Tesla will now rely more on Nvidia, AMD, and Samsung.

Samsung recently secured a $16.5 billion deal to supply AI chips for Tesla’s self-driving cars, robots, and data centres. Musk said Samsung’s Texas factory will produce Tesla’s AI6 chips, with AI5 chips to be made in 2026.

Musk suggested that combining AI5 and AI6 chips could form a ‘Dojo 3’ system, while Dojo 2 would not launch. The shutdown comes as Tesla restructures, with executive exits, job cuts, and renewed focus on AI integration across Musk’s companies.

Would you like to learn more about AI, tech, and digital diplomacy? If so, ask our Diplo chatbot!

GitHub CEO says developers will manage AI agents

GitHub’s CEO, Thomas Dohmke, envisions a future where developers no longer write code by hand but oversee AI agents that generate it. He highlights that many developers already use AI tools to assist with coding tasks.

Early adoption began with debugging, boilerplate and code snippets, and evolved into collaborative brainstorming and iterative prompting with AI. Developers are now learning to treat AI tools like partners and guide their ‘thought processes’.

According to interviews with 22 developers, half expect AI to write around 90 percent of their code within two years, while the rest foresee that happening within five. The shift is seen as a change from writing to verifying and refining AI-generated work.

Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot!

Sam Altman praises rapid AI adoption in India

OpenAI’s new GPT‑5 model has been unveiled, and the company offers it free to all users. Three model versions, gpt‑5, gpt‑5‑mini and gpt‑5‑nano, offer developers a balance of performance, cost and latency.

CEO Sam Altman applauded India’s rapid AI adoption and hinted that India, currently OpenAI’s second‑largest market, may soon become the largest. A visit to India is planned for September.

The new GPT‑5 achieves a level of expertise akin to a PhD‑level professional and is described as a meaningful step towards AGI. OpenAI intends to make the model notably accessible through its free tier.

Head of ChatGPT Nick Turley noted that GPT‑5 significantly enhances understanding across more than twelve Indian languages, reinforcing India as a key market for localisation.

Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot!

GPT-5 launch sparks backlash as OpenAI removes ChatGPT model choice

OpenAI has launched GPT-5, replacing previous ChatGPT models and removing the model picker option. CEO Sam Altman called it a PhD-level AI, claiming improvements in reasoning, writing, coding, accuracy, and health-related queries, with fewer hallucinations. The rollout followed right after the announcement.

GPT-5 includes both an efficient and a reasoning model, but users no longer choose which to engage, OpenAI’s system automatically routes queries. The change has frustrated many, as favourite models like GPT-4o and o3 are no longer available.

Users on social media and forums complain that GPT-5 gives shorter, less engaging answers and has less personality. Some say the model ignores instructions, gets basic things wrong, and is slower despite not running in ‘thinking mode’.

Several users allege OpenAI shortened responses deliberately to reduce costs, removing emotional intelligence to discourage casual chatting. Critics believe the move could result in lost subscriptions despite efficiency gains.

Others describe GPT-5 as more organised but clipped in tone, with no clear quality improvement over earlier models. The loss of previous models has left some feeling that the upgrade is a downgrade, with one user saying it feels like ‘watching a close friend die’.

Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot!

Microsoft Copilot launches GPT-5 Smart Mode for users

Microsoft has upgraded all its Copilot tools with OpenAI’s latest GPT-5 model, now available through a new Smart Mode.

The update brings enhanced speed and better performance across Microsoft 365 Copilot, Azure AI Foundry, GitHub Copilot, and the Windows AI chatbot pre-installed on PCs.

Users can try the Smart Mode via the Copilot website, which can be selected from a model selector below the text input. Notably, a Microsoft account is not required to use this new mode.

The Smart Mode rollout is gradual, and while not yet visible in the Windows Copilot app for everyone, early users have reported that GPT-5 delivers faster and more accurate responses compared to earlier models.

Unlike the free ChatGPT tier that limits queries and switches automatically to older versions after usage caps, Copilot offers fewer restrictions, allowing more extensive use.

Microsoft’s AI Red Team tested GPT-5 thoroughly before release, highlighting its strong safety profile and improved reasoning capabilities.

Microsoft 365 Copilot benefits from GPT-5’s advanced abilities to answer complex questions, maintain coherence over long conversations, assist with email replies, and analyse documents effectively.

GitHub Copilot has also been enhanced with GPT-5, significantly improving reasoning, code quality, and overall user experience. The upgrade positions Microsoft’s AI tools to support developers and professionals across multiple platforms better.

Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot!

GPT-5 launches with ‘PhD-level performance’

OpenAI has unveiled GPT-5, the latest generation of its widely used ChatGPT tool, offering what CEO Sam Altman described as a ‘huge improvement’ in capability.

Now free to all users, the model builds on previous versions but stops short of the human-like reasoning associated with accurate artificial general intelligence.

Altman compared the leap in performance to ‘talking to a PhD-level expert’ instead of a student.

While GPT-5 does not learn continuously from new experiences, it is designed to excel in coding, writing, healthcare and other specialist areas.

Industry observers say the release underscores the rapid acceleration in AI, with rivals such as Google, Meta, Microsoft, Amazon, and Elon Musk’s xAI investing heavily in the race. Chinese startup DeepSeek has also drawn attention for producing powerful models using less costly chips.

OpenAI has emphasised GPT-5’s safety features, with its research team training the system to avoid deception and prevent harmful outputs.

Alongside the flagship release, the company launched two open-weight models that can be freely downloaded and modified, a move seen as both a nod to its nonprofit origins and a challenge to competitors’ open-source offerings.

Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot!

Autonomous AI coding tool Jules now publicly available from Google

Google has released its autonomous coding agent Jules for free public use, offering AI-powered code generation, debugging, and optimisation. Built on the Gemini 2.5 Pro model, the tool completed a successful beta phase before entering general availability with both free and paid plans.

Jules supports a range of users, from developers to non-technical staff, automating tasks like building features or integrating APIs. The free version allows 15 tasks per day, while the Pro tier significantly raises the limits, providing access to more powerful tools.

Beta testing began in May 2025 and saw Jules process hundreds of thousands of tasks. Its new interface now includes visual explanations and bug fixes, refining usability. Integrated with GitHub and Gemini CLI, Jules can suggest optimisations, write tests, and even provide audio summaries.

Google positions Jules as a step beyond traditional code assistants by enabling autonomy. However, former researchers warn that oversight remains essential to avoid misuse, especially in sensitive systems where AI errors could be costly.

While its free tier may appeal to startups and hobbyists, concerns over code originality and job displacement persist. Nonetheless, Jules could reshape development workflows and lower barriers to coding for a much broader user base.

Would you like to learn more about AI, tech, and digital diplomacy? If so, ask our Diplo chatbot!

Scientists use AI to teach drones to program themselves

A computer scientist has shown that robots can now write the brains of other robots, thanks to generative AI.

Professor Peter Burke from the University of California, Irvine, has demonstrated a drone capable of creating and hosting its own control system using AI-written code, significantly reducing the time usually needed to build such infrastructure.

The project used several AI models and coding tools to prompt the creation of a real-time, web-based command centre hosted on the drone itself. The final system, which runs on a Raspberry Pi Zero 2 W, allows the drone to operate independently while remaining accessible over the internet.

Unlike traditional systems, where ground control is handled externally, the drone manages its own mission planning and navigation through a built-in AI-generated website.

Burke’s team used tools such as Claude, Gemini, ChatGPT, Cursor, and Windsurf to build the system across several sprints. Despite context limitations in each model, the final version was completed in just over 100 hours, around twenty times faster than a previous project of similar complexity.

The final codebase consisted of 10,000 lines and included everything from flight commands to map-based interaction and GPS tracking.

Although the technology shows promising potential in fields like aerial imagery and spatial AI, experts have raised safety concerns.

While a manual override system was included in the experiment, the ability for robots to self-generate control logic introduces new ethical and operational challenges, especially as such systems evolve to operate in unpredictable environments.

Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot!

Endex brings AI to Excel with OpenAI Startup Fund support

Endex.ai has secured $14 million in funding to bring AI directly into Microsoft Excel. The funding round was led by the OpenAI Startup Fund, marking a significant moment for traditional tools in the business world.

Founded in 2022 by Tarun Amasa and Kevin Yang, the startup has spent the past year collaborating quietly with financial institutions to refine its product.

Now available to the public through limited invites, the tool embeds itself within Excel and helps users manage tasks like financial modelling, data cleanup, and detailed analysis (without switching applications).

Unlike broader AI tools, Endex has been designed specifically for finance professionals. It understands financial terminology, adapts to user habits, and references trusted data sources such as SEC filings, CapIQ, and earnings reports.

The company describes its product as Excel-native, aiming to enhance rather than replace a tool already deeply integrated into finance work.

With the new funding, Endex plans to expand development and scale its reach. The AI agent already works on both Mac and Windows, and its frictionless interface is proving attractive in a field where saving time and improving accuracy can make a substantial difference.

Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot!