Lloyds Banking Group to hire 300 tech experts to work on AI ↗
Lloyds is adding 300 tech specialists to build out agentic AI, with recruits joining a wider 1,000-person AI team. The bank wants the systems working on fraud, HR document search, and more personal online banking tools.
The tension, plainly, is that this is hiring news wrapped around automation anxiety. Lloyds says AI has already brought tens of millions in gains, while leaving open the possibility of job cuts as roles continue to shift.
The Atlantic created a searchable database of the music used to train AI ↗
A searchable database now lets people look through music datasets used for AI training. Some are vast - one has 12 million tracks, another has 9 million, plus smaller collections still topping 100,000 songs.
The uncomfortable bit: tracks can be freely listed online while still not being cleared for commercial training. It is one of those “public but not permissionless” tangles that keeps turning the AI-copyright crank.
In the Weights is your new AI-centric vanity search ↗
In the Weights turns ego-searching into an AI-era scoreboard. Instead of asking Google who you are, it asks models like GPT, Claude, Gemini, Grok, and Llama what they “remember” about a name.
The site gives people a strength score and shows where models hallucinate. Peculiarly fun, faintly cursed, and maybe a tiny mirror maze for the whole “am I in the training data?” obsession.
Signal’s Meredith Whittaker wants you to remember that AI chatbots ‘are not your friends’ ↗
Meredith Whittaker warned against treating chatbots as companions or trusted thinking partners. Her point was blunt: these systems are not conscious, not friends, and not private confidants.
She also pushed back on agentic assistants that need access to chats, credit cards, browsers, calendars, and messages. In Signal’s world, that kind of convenience starts looking a lot like a backdoor.
Nobel laureate John Jumper is leaving DeepMind for rival Anthropic ↗
John Jumper, known for his AlphaFold work, is leaving Google DeepMind for Anthropic. That is not just a personnel move - it is a bright neon sign that top AI talent is still moving hard between frontier labs.
The timing lands alongside other high-profile DeepMind departures. So, yes, the lab still has enormous prestige, but the gravity well around Anthropic is looking pretty serious.
AI is making crypto security cheaper, faster and harder to ignore ↗
AI security tools like Mythos could push smart-contract auditing toward cheaper, continuous review. That matters because crypto audits have often been expensive, slow, and treated as optional by smaller teams.
Researchers still warn that AI will not replace human judgment. Plenty of crypto disasters come from compromised credentials, social engineering, or brittle operations - not just buggy code.
Japan to Enhance Global Cooperation on AI Risks ↗
Japan released a draft update to its Artificial Intelligence Basic Plan, putting more emphasis on cooperation with foreign governments and AI developers. The focus is misuse, cyberattacks, and fast-moving model risks.
It is a policy reset made only months after the plan was first formed. That says the quiet part out loud: AI governance is aging like milk in a heatwave.
New global order: AI CEOs as heads of nation-states ↗
At the G7 summit, AI CEOs were seated with world leaders in a scene that felt, plainly, a bit sci-fi council chamber. OpenAI, Google DeepMind, Anthropic, and Salesforce leaders were treated as central actors in global security and governance.
The core issue is power. Governments want dependable AI partners, but they also do not want private labs deciding the future by themselves - or so everyone says, until the next crisis lands.
FAQ
Why is Lloyds Banking Group hiring more AI specialists?
Lloyds is hiring 300 tech specialists to expand its work on agentic AI, adding to a wider AI team of about 1,000 people. The bank wants to use these systems in areas such as fraud detection, HR document search, and more personalised online banking. The article also notes the tension between AI-driven gains and uncertainty over how automation may alter jobs.
Why does the AI music training database matter?
The searchable music database matters because it lets people inspect tracks included in datasets used for AI training. Some datasets are vast, containing millions of tracks. The central issue is that music can be publicly listed online without necessarily being cleared for commercial AI training, leaving copyright and permission questions unresolved.
What is “In the Weights” and how does it relate to AI search?
In the Weights is described as an AI-era vanity search tool. Instead of showing standard web results, it asks models such as GPT, Claude, Gemini, Grok, and Llama what they appear to “remember” about a name. It then provides a strength score and shows where models hallucinate, making it a playful but uneasy reflection of training-data curiosity.
Why are privacy advocates warning about AI chatbots?
Signal’s Meredith Whittaker warns that chatbots should not be treated as friends, conscious beings, or private confidants. Her concern is especially sharp around agentic assistants that need access to messages, calendars, browsers, credit cards, and chats. In privacy-focused systems, that kind of convenience can begin to look like a serious security weakness.
What does John Jumper’s move from DeepMind to Anthropic suggest?
John Jumper’s move from Google DeepMind to Anthropic suggests that top AI talent is still moving aggressively between frontier labs. Because Jumper is known for AlphaFold, the departure is more than routine hiring news. The article frames it as another sign that Anthropic has become a serious talent magnet, even while DeepMind remains highly prestigious.