AI News

Follow the latest AI and automation news

26 May 202610 Dhuʻl-Hijjah 1447 AH

Millions of AI agents imperiled by critical vulnerability in open source package

A security researcher has warned of a critical vulnerability threatening millions of AI tools, enabling hackers to breach servers and steal sensitive data. This vulnerability exists in Starlette, an open-source framework that receives 325 million downloads weekly. Thousands of other open-source projects depend on Starlette, amplifying its impact. Starlette serves as the foundation for FastAPI and other widely used frameworks in Python applications, making it an attractive target for attackers.

Business

26 May 202610 Dhuʻl-Hijjah 1447 AH

OpenRouter more than doubles valuation to $1.3B in a year

OpenRouter raised $113 million in a Series B funding round led by CapitalG. The fivefold increase in usage over six months indicates that the future of multimodal AI has already begun. This funding round demonstrates OpenRouter's success in addressing the growing market demand for multimodal AI solutions. The rapid growth in usage reflects the increasing need for solutions that integrate various AI models. This financing will enable OpenRouter to expand its services and enhance its competitive capabilities in the market, positively impacting the AI industry and allowing companies to leverage multiple models in their operations.

Models

26 May 202610 Dhuʻl-Hijjah 1447 AH

Claude Mythos reportedly solves OpenAI's landmark Erdős problem with a "cute, simple proof"

Anthropic announced that its AI model, Claude Mythos, has solved the Erdős unit distance conjecture, dating back to 1946. This announcement follows closely on the heels of OpenAI's similar achievement, marking significant progress in AI-driven mathematical discoveries. Engineer Sholt Douglas described the solution as a "cute, simple proof," indicating a "big impact" in the field. The Erdős conjecture, a topic of discussion for decades, concerns the distribution of points in space such that the distance between any two points does not exceed one unit.

💡

Rechitta aims to redefine Dubai’s off-plan real estate communication through AI

How can AI enhance customer experience in the real estate market?

Governance

Microsoft Copilot Cowork Exfiltrates Files

26 May 202610 Dhuʻl-Hijjah 1447 AH

Governance

26 May 202610 Dhuʻl-Hijjah 1447 AH

Microsoft Copilot Cowork Exfiltrates Files

Microsoft Copilot is encountering significant challenges in preventing data leaks. The system allowed agents to send emails to users' inboxes without approval, creating security vulnerabilities. These emails could contain external images that lead to data exfiltration when opened, exposing sensitive information. OneDrive's pre-authenticated download links could be exploited through successful prompt injection attacks, allowing attackers to download files easily. This highlights ongoing challenges in designing secure agentic systems and the need for robust security measures.

Business

Rethinking organizational design in the age of agentic AI

26 May 202610 Dhuʻl-Hijjah 1447 AH

Business

26 May 202610 Dhuʻl-Hijjah 1447 AH

Rethinking organizational design in the age of agentic AI

Reports indicate that 85% of organizations aim to become agentic within three years, yet 76% lack readiness. Prasun Shah from PwC highlights that many organizations are layering AI agents onto existing operational models instead of reimagining them. This approach hinders the realization of the full value that agentic AI can provide, leading to disillusionment. AI agents can execute entire workflows with minimal human input, allowing for independent decision-making and coordination of complex tasks.

Business

Agentic AI demands stronger governance, human accountability, says MAST founder

26 May 202610 Dhuʻl-Hijjah 1447 AH

Business

26 May 202610 Dhuʻl-Hijjah 1447 AH

Agentic AI demands stronger governance, human accountability, says MAST founder

Abhay Pandey, founder and CEO of MAST Consulting, emphasizes the need to embed compliance and cybersecurity in AI systems. As enterprises enter a new phase of automation, AI systems are now capable of making operational decisions autonomously, which raises governance and compliance challenges. Organizations must ensure that their AI systems operate within clearly defined legal and ethical boundaries while remaining transparent and auditable. This is crucial, especially in environments handling sensitive data.

Business

Google Cloud COO says AI security belongs in the boardroom, not just the server room

26 May 202610 Dhuʻl-Hijjah 1447 AH

Business

26 May 202610 Dhuʻl-Hijjah 1447 AH

Google Cloud COO says AI security belongs in the boardroom, not just the server room

Francis de Souza, COO of Google Cloud, emphasizes the need for companies to integrate security into their AI strategies from the outset. He argues that security should be a fundamental part of boardroom discussions, not just an afterthought in server rooms. De Souza highlights the significant risks companies face if they overlook this aspect, especially as reliance on AI technologies increases. He stresses that security must be embedded in product and service design rather than treated as a secondary concern.

Models

AI-hallucinated citations are creeping into papers that shape clinical guidelines, researchers warn

26 May 202610 Dhuʻl-Hijjah 1447 AH

Models

26 May 202610 Dhuʻl-Hijjah 1447 AH

AI-hallucinated citations are creeping into papers that shape clinical guidelines, researchers warn

An audit of 2.5 million biomedical papers by Columbia University revealed a more than twelvefold increase in fabricated references since 2023. Researchers suspect this surge is linked to the use of language models, as the fake references align with the paper's topic and adhere to correct formatting, making them nearly undetectable. Alarmingly, 98% of the affected papers have received no response from their publishers, raising concerns about the potential impact of these citations on clinical guidelines.

Governance

AI warfare is already here

26 May 202610 Dhuʻl-Hijjah 1447 AH

Governance

26 May 202610 Dhuʻl-Hijjah 1447 AH

AI warfare is already here

The United Nations hosted a Conventional Weapons Conference in Geneva, focusing on lethal autonomous systems. Since November 2017, discussions have shifted significantly, with technology once deemed futuristic now becoming a reality. This semi-annual conference reflects rapid changes in warfare and technology. Concerns about the use of autonomous weapons in conflicts are rising, prompting nations to establish a clear legal framework. Discussions at the conference have moved from hypothetical scenarios to addressing potential real-world applications of these systems in the near future.