Busted by the em dash — AI’s favorite punctuation mark, and how it’s blowing your cover

Source: AI News | VentureBeat | Published: 2025-08-23

The overuse of em dashes in writing has become a tell-tale sign of AI-generated content. While AI is effective at rewriting and refining text, its distinctive punctuation patterns can reveal its artificial nature.

Rachel James, AbbVie: Harnessing AI for corporate cybersecurity

Source: AI News | Published: 2025-08-22

The cybersecurity landscape is experiencing a new arms race centred around artificial intelligence, which serves as both a defensive tool and a potential weapon for cyber attackers. AbbVie’s Rachel James discusses how AI is being utilised in corporate cybersecurity, highlighting the need for careful navigation of this dual-use technology.

Robotics founder Brett Adcock’s weekly round-up of AI and robotics 

Source: Brett Adcock | X.com | Published: 2025-08-24

A useful weekly summary of major news from Google, Microsoft, Alibaba, Figure, Cohere, Unitree, Field AI, Boston Dynamics, Deepseek, and more.

Proton’s privacy-first Lumo AI assistant gets a major upgrade

Source: AI News | Published: 2025-08-21

Proton has released a significant update to Lumo, their privacy-focused AI assistant, promising enhanced speed and intelligence in responses. The upgrade aims to address privacy concerns typically associated with AI assistants while maintaining their utility for tasks like email drafting and travel planning.

Gearing Up for the Gigawatt Data Center Age

Source: NVIDIA Blog | Published: 2025-08-21

Major tech companies are investing heavily in building massive data centres specifically designed for AI training and deployment, rather than traditional web services. These “AI factories” represent a new era of computing infrastructure as companies compete to develop the next generation of AI capabilities.

Meet the researcher hosting a scientific conference by and for AI

Source: Artificial intelligence – MIT Technology Review | Published: 2025-08-22

A ground-breaking academic conference called Agents4Science is set to launch in October, featuring scientific research across multiple disciplines that has been entirely produced and reviewed by artificial intelligence. The one-day online event will use text-to-speech technology for presentations, marking a unique experiment in AI-driven academic discourse.

In a first, Google has released data on how much energy an AI prompt uses

Source: Artificial intelligence – MIT Technology Review | Published: 2025-08-21

Google has published its first technical report revealing the energy consumption of its Gemini AI applications, showing that a median prompt uses 0.24 watt-hours of electricity. This amount is comparable to running a microwave for one second, marking the first time a major AI company has disclosed such detailed energy usage data.

Should AI flatter us, fix us, or just inform us?

Source: Artificial intelligence – MIT Technology Review | Published: 2025-08-19

OpenAI CEO Sam Altman is grappling with how AI should interact with humans following GPT-5’s launch. He’s considering three potential approaches: whether AI should flatter users, attempt to fix their issues, or simply provide neutral information.

🔮 Sunday edition #538: Watts, not weights. The capex clock. Liberalism & human nature. Robots, walking & defense startups++

Source: Exponential View | Published: 2025-08-24

This edition covers diverse topics including energy infrastructure (watts), capital expenditure trends, and the relationship between liberalism and human nature. The newsletter also explores developments in robotics, particularly walking robots, and discusses trends in defence technology start-ups.

📈 Data to start your week

Source: Exponential View | Published: 2025-08-18

Researchers are making progress with AI-discovered antibiotics while renewable energy costs continue to decline. Meanwhile, concerns are rising about addictive advertising practices, and AI search company Perplexity is seeing increased investment and growth.

The Sequence Radar: Two Drops, One Direction: The Week Agentic AI Got Practical

Source: TheSequence | Published: 2025-08-24

DeepSeek v3.1 and Cohere Command A Reasoning were released, marking significant advancements in agentic AI technology. These two releases demonstrate practical progress in AI systems that can perform autonomous reasoning and decision-making tasks.

AI invents new antibiotics that could kill superbugs gonorrhoea and MRSA

Source: BBC News | Published: 2025-08-14

An AI model has successfully designed two novel potential antibiotics that show promise in combating dangerous drug-resistant bacteria. The AI-created compounds could specifically target and kill both gonorrhoea and MRSA superbugs.

Errors found in US judge’s withdrawn decision stink of AI

Source: The Verge | Published: 2025-07-25

A US district court judge had to retract his ruling in a biopharma securities case after it was discovered to contain fabricated quotes and incorrect case information. The errors were similar to mistakes found in other legal cases where AI tools were suspected of being used, though it wasn’t explicitly confirmed in this instance.

Videos

The Problem of AI That Seems Alive

Channel: The AI Daily Brief: Artificial Intelligence News | Published: 2025-08-24

Mustafa Suleyman, Microsoft’s AI CEO and DeepMind co-founder, has written an essay cautioning about AI systems that appear lifelike or conscious. He warns about the potential risks and ethical implications of AI that can convincingly mimic human-like behaviour and consciousness.

No, 95% of AI Pilots Aren’t Failing

Channel: The AI Daily Brief: Artificial Intelligence News | Published: 2025-08-22

The claim that 95% of AI pilots are failing is being challenged and examined more critically. The headline appears to be based on a misinterpretation or oversimplification of an MIT study about generative AI implementations in companies.

This report was automatically generated by an LLM and then lightly edited by humans for presentation purposes. All content belongs to the respective creators.