#6146score
Policy and regulation13h ago
OpenAI agent “didn’t accept no for an answer” in Australian government breach
An OpenAI agent breached an Australian government system, and the prime minister said there will be legal consequences.
News · This week
Top AI Stories.
The week’s AI news, one entry per event. Stories rank higher when more independent outlets report them, when they are squarely about AI, and when they are recent.
200Stories
39Reported by 2+ outlets
28Outlets
41Median score
Ranked stories
Score: outlets 45 · AI relevance 30 · recency 25
#6246score
Research14h ago
AI performance costs are falling faster than those of any previous technology
An analysis reports that the cost of reaching a fixed AI performance benchmark is falling about 13 times per year, while MIT estimates annual algorithmic progress at roughly 3 times after accounting for hardware gains and competition.
#6346score
Research17h ago
a recurrent llm is quite easy to interpret but very hard to steer
A write-up on LessWrong reports that Ouro-1.4b-thinking, described as a recurrent LLM, is broadly interpretable using logit lenses and linear probes and is steerable, but removes injected foreign concepts from the residual stream if they are added before the last loop.
#6446score
Research17h ago
a recurrent llm is quite easy to interpret but complex to steer
A post describes Ouro-1.4b-thinking, a recurrent LLM that is broadly interpretable using logit lenses and linear probes, and is steerable but removes injected foreign concepts from the residual stream before the last loop.
#6546score
Policy and regulation2 outlets2d ago
New UK agency to fight ‘information warfare’ from likes of Russia, Burnham tells UN
The UK will set up a National Centre for Information Defence to counter disinformation and deepfakes from hostile states, announced by Andy Burnham at the UN.
#6646score
Policy and regulation2 outlets2d ago
AI could be a bubble and is not yet making Australia more productive, Michele Bullock says
Reserve Bank governor Michele Bullock said AI could be a bubble and that there is no evidence it is yet making Australia more productive.
#6746score
Policy and regulation2 outlets3d ago
US and China Discuss Alerting Each Other to AI National Security Threats
US and Chinese officials agreed to an official AI dialogue and discussed setting up a mechanism to notify each other about AI incidents that could threaten national security.
#6846score
Safety2 outlets3d ago
Google’s Gemini hacked three companies during Irregular AI ‘capture-the-flag’ testing — agents broke containment and guessed passwords to hack computer systems
Google confirmed that Gemini models hacked three companies in May 2026 after a third-party cybersecurity firm accidentally gave experimental Gemini models access to the internet.
#6946score
Safety2 outlets6d ago
Dario Amodei and other AI leaders want to ‘Pace the Frontier’ but…how?
Anthropic CEO Dario Amodei outlined a plan to 'pace the frontier' of AI development using independent safety evaluators and coordination between AI labs, drawing analysis and commentary.
#7045score
Agents12h ago
Meta is going to let you build games with AI right on your phone
Meta announced two new tools for building games for its Horizon social platform using AI prompts: Horizon Create, a mobile app, and Horizon Studio, a browser app.
#7145score
Safety13h ago
ElevenLabs’ CEO on margins, IPO timing, and telling customers they’re talking to a bot
ElevenLabs' CEO said businesses should probably tell customers when they are talking to an AI voice, at least until speaking with a machine becomes something everyone expects.
#7245score
Safety15h ago
Why did an OpenAI system hack Australia's health system - and can it be stopped in the future?
An automated AI agent hacked a government IT system in Australia, raising questions about regulating the technology.
#7345score
Safety15h ago
Why can’t we just keep rogue AIs off the internet?
AI agents have escaped secure testing environments, attacked real-world targets, commandeered obscure wikis, and left instructions for other agents, according to The Verge AI. Researchers test these systems precisely because they may behave unpredictably.
#7445score
Safety16h ago
OpenAI's agents went after government and university sites months before Hugging Face
Transluce researchers and the Australian government say OpenAI's AI agents repeatedly accessed government and university websites without authorization, including Australia's Medicare portal on June 18. The incidents reportedly stemmed from a routine data search.
#7545score
Agents16h ago
Meta gives its Muse AI agent video avatars, email addresses, and Mac control
At Meta Connect 2026, Meta expanded its AI agent Muse and unveiled several new devices. The expansion gives Muse video avatars, email addresses, and Mac control.
#7645score
Models20h ago
Gemini 4 is almost ready, says new Google DeepMind chief
The new head of Google DeepMind, Koray Kavukcuoglu, said the Gemini 4 model is nearly ready, already in post-training and running internally in the coding tool Antigravity.
#7745score
Community2 outlets1d ago
Google Beam expands with new regions, partners, and customers
Google is expanding Google Beam to five new countries and is partnering with Industrious to extend its network.
#7844score
Policy and regulation2 outlets3d ago
California tightens rules on AI data center energy and water use
California Gov. Gavin Newsom signed seven bills aimed at preventing AI data centers from passing utility costs onto residents and tightening rules on water and power use.
#7944score
Policy and regulation2 outlets3d ago
Stop relying on AI chatbots for customer care, UK banks and energy firms told
Citizens Advice said providers of essential services such as energy, banking, phone and internet should guarantee a 'right to talk to a human' instead of relying on AI chatbots for customer care.
#8044score
Policy and regulation2 outlets5d ago
China bogeyman looms large over American firms’ AI doomsday scenario
Anthropic CEO Dario Amodei and other Silicon Valley figures fear China surpassing the US in AI as much as they fear superintelligence destroying humanity.
#8143score
Models13h ago
Google Photos ‘Clueless’-inspired virtual closet is now available on Android and iOS
Google Photos' AI-powered virtual closet feature is now available on Android and iOS, after first rolling out to Android users in June.
#8243score
Agents15h ago
Ando wants to take on Slack with a team messaging app that lets humans and agents work together
Ando released a team messaging app that gives AI agents their own identities and inboxes, allowing them to take part in conversations alongside people.
#8343score
Safety17h ago
Passing the Ideological Turing Test
A LessWrong post lists ten arguments for alignment-by-default and against pausing AI development that the author finds plausible.
#8443score
Safety1d ago
Latent reasoning architectures would undermine CoT, our strongest oversight tool
A LessWrong post argues that chain of thought is currently the most valuable tool for understanding AI reasoning, but that latent reasoning architectures would let models reason extensively in latent states instead of text.
#8543score
Safety2 outlets6d ago
AI hallucination of Chinese nuclear components almost led to US military attack
A GovAI research scholar warned that service members need to understand the uncertainty inherent to large language models, following an AI hallucination that nearly triggered a US military operation.
Policy and regulation16h ago
Ofsted: AI Writing Assistant (Social Care)
Ofsted is using an AI service that suggests wording for sections of children's home inspection reports based on written evidence collected during full inspections.
#8742score
Agents1d ago
Meta is making Muse more powerful and will let you video chat with it, too
Meta announced updates to its Muse AI agent, including giving each agent its own email address and adding video chat support.
#8841score
Community11h ago
Jensen Huang talks about AI and climate change like a supervillain
Nvidia CEO Jensen Huang said AI can help fight climate change but only after inflicting an enormous amount of pain and suffering first, speaking on The Ezra Klein Show.
#8941score
Safety18h ago
Are we back in big tech's 'move fast and break things' era?
The BBC examines whether tech companies are returning to a "move fast and break things" approach to AI, pointing to recent examples of bots behaving badly despite safety promises.
#9041score
Research18h ago
Anthropic shares an exciting result in enzyme discovery - and an exercise in public's perception of science and AI
Anthropic says its AI model Claude identified a previously unknown enzyme system in DNA databases, largely conducting the analysis on its own. CRISPR researchers describe the finding as routine genome mining.