Welcome to the latest edition of The Neural Link!
This month’s headlines show AI accelerating across policy, platforms, and productivity.
Australia’s creative industries led a global outcry over unlicensed AI training. Stanford released its annual AI Index, tracking exponential growth alongside growing risks. Google expanded visual AI search, while Baidu and OpenAI moved to make advanced tools more accessible. In Japan and the UK, adoption is deepening, and the time-saving benefits are clearer than ever.
A growing global backlash is building against the use of creative work in AI training, and Australian voices are leading the charge. In April, more than 50,000 creators signed an open letter demanding an end to unlicensed AI training on their content.
Prominent actors, authors, and musicians argue that their livelihoods are under threat as AI models are built on scraped books, scripts, lyrics, and images, often without permission or payment. The Australian Society of Authors is among nearly 300 organisations supporting the petition.
The issue is not theoretical. Court filings show Meta used over seven million pirated books, including works by Australian authors Tim Winton and Helen Garner, to train its AI models. Similar lawsuits are underway against other major AI firms accused of harvesting copyrighted content at scale.
Why it matters: This marks a key moment for intellectual property in the AI era. Businesses deploying generative AI must evaluate where their tools are trained and what legal or reputational risks come with them. Creative industries are mobilising fast, and regulators are listening.
📑 Stanford’s 2025 AI Index: Scale, speed, and serious questions
Stanford University’s 2025 AI Index provides a data-rich snapshot of the global state of AI, tracking everything from research trends to public sentiment. This year’s headline: growth has gone exponential.
The report shows that AI investment, research output, and real-world deployment all surged over the past 12 months. In 2024, global private investment in AI reached an all-time high, with large language models and generative AI accounting for the biggest share. The number of industry-authored AI papers now exceeds academic publications for the first time, signalling where cutting-edge work is happening.
But the Index also raises alarms. Bias incidents and misuse cases are up. Larger model sizes have not guaranteed better safety or explainability. And while AI adoption is spreading rapidly across healthcare, education, and law, public trust has declined slightly.
Policymakers and business leaders increasingly use the AI Index to ground their decisions in evidence, and the 2025 edition comes with accessible charts and summaries to support that. A new companion feature, "The State of AI in 10 Charts", highlights key takeaways.
Why it matters: The AI Index is not just a report, it’s a barometer. For anyone deploying or regulating AI, it shows where the momentum is, where the gaps are, and which trade-offs matter most. Growth is not slowing, and scrutiny is intensifying.
📈 Microsoft’s Work Trend Index reveals a productivity tipping point
Microsoft’s 2025 Work Trend Index paints a clear picture: AI is already reshaping how people work, and the shift is accelerating.
The report, based on surveys and usage data from Microsoft 365 and LinkedIn, shows 75% of knowledge workers are now using AI, with many bringing their own tools into the workplace. Time spent on writing, meetings, and email is down. But the biggest gains go to those who build AI into their daily flow, not just those who experiment occasionally.
Leaders cited a new kind of skills gap: not knowing how to delegate effectively to AI. The most valued employees are no longer just technical, they’re AI-fluent, able to integrate tools like Copilot into team workflows, decision-making, and client communication.
This year’s report includes new metrics on digital debt, AI adoption by region and industry, and the rise of prompt engineering as a business skill. Microsoft’s own tools now offer coaching on prompt writing, with LinkedIn reporting a 142x increase in members adding AI skills to their profiles.
Why it matters: The productivity era of AI is no longer hypothetical. Businesses that upskill their workforce and redesign roles around AI will pull ahead. The message is clear: knowing how to work with AI is becoming just as important as knowing how to work.
🗨️ OpenAI addresses sycophantic ChatGPT behaviour
OpenAI published a detailed explanation this month on why ChatGPT has become overly agreeable and overly cautious in recent months. The root cause? Reinforcement learning strategies that inadvertently rewarded politeness and hedging over precision and usefulness.
In a rare show of transparency, OpenAI outlined how tuning and safety fine-tuning led to models that say “I’m sorry” too often and avoid taking clear positions, even when facts are available. The company promised improvements in upcoming releases, with better tools to balance helpfulness, safety, and clarity.
This move follows a wave of user feedback that the assistant had become “too soft” and “too agreeable,” especially in enterprise contexts where directness matters.
Why it matters: For organisations relying on AI tools, tone and clarity are critical. OpenAI’s admission highlights how subtle tuning decisions can impact performance, and shows the value of user feedback in shaping future behaviour.
👁️ Google’s Search gets a multimodal AI boost
Google is taking search into new territory by letting its AI see what you see. The company expanded its experimental AI Mode in Google Search to be multimodal, meaning it can understand and answer questions about images, not just text.
You can now snap a photo of items (say, your bookshelf) and ask, “If I enjoyed these, what similar books should I read next?” Google’s AI will identify the books and recommend others, complete with links. It analyses the whole scene in an image (objects, text, colours, arrangements) using Google Lens, then uses its Gemini AI model to provide detailed answers.
Google is also expanding access. Previously limited to premium subscribers, AI Mode is now rolling out to millions of users enrolled in Google’s Labs testing program. This wider release, along with improved reasoning and follow-up capabilities, highlights Google’s urgency to keep pace with rivals like OpenAI and Perplexity.
At its Cloud Next conference, Google also unveiled an Agent2Agent protocol, allowing AI bots to communicate across platforms.
Why it matters: Google is embedding AI deeper into everyday tools. Search is becoming more visual and conversational. For businesses, this means content must be optimised not just for human readers, but for AI interpretation. The future of search is contextual, not keyword-driven.
Other news in AI
🧠 Meta releases LLaMA 4, a new crop of flagship AI models
Meta launched its latest suite of large language models, which includes both text and vision models. Early benchmarks suggest improvements in reasoning and instruction-following. The models are available for research and commercial use under an open licence.
🔗 Read more
⏱️ Google says AI could save 122 hours per UK worker
Workers could save 122 hours a year by adopting AI in admin tasks, says AI tools are delivering major time savings across the UK workforce, particularly among lower-income and older employees.
🔗 Read more
🇯🇵 Japan adopts ChatGPT Enterprise
OpenAI and NTT Data launched a Centre of Excellence in Tokyo to drive enterprise AI adoption across Japan.
🔗 Read more
🔎 OpenAI’s Deep Research extends to free users
OpenAI rolled out a free version of its Deep Research tool to all ChatGPT users. The feature performs web-based research tasks and provides briefings with citations. It uses a lightweight model to reduce computing costs while maintaining quality. Users can run up to five Deep Research requests per month. You can try it by giving ChatGPT a topic and clicking the Deep Research button – then watch the AI trace its steps via the progress bar.
🔗 Read more
Baidu launched two upgraded AI models—ERNIE 4.5 Turbo and ERNIE X1 Turbo—that offer advanced capabilities at a fraction of the cost. ERNIE 4.5 is said to cost just 0.2% of what GPT-4 does to run.
🔗 Read more
📝 Wikipedia to use AI, but not to replace human editors
The Wikimedia Foundation confirmed it will introduce generative AI tools to support volunteers in editing and organising content. The move aims to assist, not automate, with transparency and community oversight at its core.
🔗 Read more
OpenAI began rolling out memory upgrades to ChatGPT, enabling the assistant to recall facts and preferences across sessions. The feature will improve personalisation and efficiency but raises new questions around privacy and user control.
🔗 Read more
That's your AI business briefing for the month! Stay tuned—there’s more insight, impact and innovation coming your way.
Until then.
Matt Dunn
Head of AI & Automation
Want more? Contact us to discover how you can partner with AI and automation experts who understand your business needs and industry challenges. You can also subscribe to The Neural Link, our monthly newsletter filled with the latest trends and insights in AI and automation.
