CAMIA privacy attack reveals what AI models memorise

Source: AI News | Published: 2025-09-26

Researchers have developed a new, more effective privacy attack called CAMIA (Context-Aware Membership Inference Attack). The method, created by researchers from Brave and the National University of Singapore, reveals what AI models have “memorised” by determining if a specific person’s data was used in the model’s training set. This new technique exposes significant privacy vulnerabilities in artificial intelligence.

OpenAI, Oracle, SoftBank To Invest $400bn In Five AI Data Centres

Source: Technology News | Silicon UK | Published: 2025-09-25

OpenAI, Oracle, and SoftBank are partnering on a massive $400 billion investment to build five “Stargate” AI data centres across the United States. This significant commitment to AI infrastructure comes as some market analysts are beginning to question the sustainability of the current boom, raising concerns about a potential AI bubble.

AI Now Co-ED Amba Kak Gives Remarks Before the UN General Assembly on AI Governance

Source: News Archives – AI Now Institute | Published: 2025-09-26

Amba Kak, Co-Executive Director of the AI Now Institute, delivered remarks at the UN General Assembly during a high-level meeting on AI governance. Her speech was part of the official launch of the Global Dialogue on AI Governance, a new multistakeholder initiative. This event marks a significant step in fostering international cooperation on shaping AI policy.

Oxford Robotics Institute director discusses the truth about AI and robotics

Source: The Robot Report | Published: 2025-09-28

Professor Nick Hawes, director of the Oxford Robotics Institute, is offering his perspective on the current state of artificial intelligence and robotics. He discusses the reality of where the technology stands today, separating it from the hype. The University of Oxford professor also explores the future potential of these fields.

🔮 Exponential View #543: AI vs youth; flying wind turbines; quantum bond trading, OpenAI’s movie & gene therapy ftw++

Source: Exponential View | Published: 2025-09-27

The latest Exponential View newsletter helps leaders understand the future of AI and technology. This issue explores a diverse range of topics, including the impact of AI on youth, innovations like flying wind turbines and quantum bond trading, and major developments in gene therapy.

The Sequence Radar #727: Qwen’s One‑Week Gauntlet

Source: TheSequence | Published: 2025-09-28

Alibaba’s Qwen is undergoing a period of rapid expansion, releasing a suite of new AI models at an incredible speed. This intense push, dubbed a “one-week gauntlet,” highlights the Chinese company’s aggressive development and release pace in the competitive AI landscape.

Videos

Stop Blaming AI For Workslop

Channel: The AI Daily Brief: Artificial Intelligence News | Published: 2025-09-28

“Workslop” is the emerging term for AI-generated content that appears polished but lacks substance and quality. The author argues that instead of blaming the technology, we should focus on the human decisions and incentives that lead to the creation of this low-effort output. AI is merely the tool, not the root cause of the problem.

How ChatGPT Pulse Changes AI

Channel: The AI Daily Brief: Artificial Intelligence News | Published: 2025-09-27

OpenAI has unveiled ChatGPT Pulse, a new AI agent designed to work proactively in the background. This marks a significant change from current AI models, shifting them from reactive tools that answer prompts to continuous assistants that provide daily, anticipatory support.

Could AI Wipe Out 1/5 of Public Companies?

Channel: The AI Daily Brief: Artificial Intelligence News | Published: 2025-09-26

A new analysis suggests that the rapid advancement of artificial intelligence could cause a massive corporate disruption, potentially leading to the failure of one-fifth of all public companies. This forecast, presented in content sponsored by KPMG, highlights the urgent need for businesses to adopt AI strategies to remain competitive.

OpenAI Tests if GPT-5 Can Automate Your Job – 4 Unexpected Findings

Channel: AI Explained | Published: 2025-09-26

A new OpenAI report “GDPval: Evaluating AI model performance on real-world economically valuable tasks” investigates the potential for a next-generation AI, like a hypothetical GPT-5, to automate jobs. The research revealed four unexpected findings from its tests. This study provides the clearest look yet at how advanced AI could impact the workforce by 2025.

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