Artificial intelligence

Hackers exploit AI: The hidden dangers of open-source models

As AI adoption grows, security experts warn that malicious actors are finding new ways to exploit vulnerabilities in open-source models.

Yuval Fernbach, CTO of machine learning operations at JFrog, notes that hackers are increasingly embedding harmful code within AI models, making it easier to steal information, manipulate outputs, or disrupt services.

A recent study by JFrog and Hugging Face found that of over one million AI models analyzed, 400 contained malicious code—roughly a 1% chance of encountering a tainted model.

However, the risk has escalated: while the number of available AI models has tripled, attacks have increased sevenfold.

The widespread use of open-source models, often chosen over costly proprietary alternatives, exacerbates security concerns.

Many companies lack proper oversight, with 58% of surveyed firms admitting to having no formal policy for vetting AI models. Meanwhile, banks and other industries worry that AI’s rapid evolution outpaces their ability to implement safeguards.

With agentic AI poised to automate decision-making, businesses face an urgent need to strengthen AI security measures before vulnerabilities lead to significant financial and operational consequences.

For more information on these topics, visit diplomacy.edu.

AI technology sparks debate in Hollywood

Hollywood is grappling with AI’s increasing role in filmmaking, with executives, actors, and developers exploring the technology’s potential. At a recent event, industry leaders discussed AI-generated video, heralded as the biggest breakthrough since the advent of sound in cinema.

Despite its growing presence, AI’s impact remains controversial, especially after recent strikes from actors and writers seeking protection from AI exploitation.

AI technology is making its way into movies and TV shows, with Oscar-nominated films like Emilia Perez and The Brutalist using AI for voice alterations and actor de-aging. AI’s capacity to generate scripts, animation, and even actors has led to fears of job displacement, particularly for background actors.

However, proponents like Bryn Mooser of Moonvalley argue that AI can empower filmmakers, especially independent creators, to produce high-quality content at a fraction of traditional costs.

While Hollywood is still divided on AI’s potential, several tech companies, including OpenAI and Google, are lobbying for AI models to access copyrighted art to fuel their development, claiming it’s vital for national security.

The push has met resistance from filmmakers who fear it could undermine the creative industry, which provides millions of jobs. Despite the opposition, AI’s role in filmmaking is rapidly expanding, and its future remains uncertain.

Some in the industry believe AI, if used correctly, can enhance creativity by allowing filmmakers to create worlds and narratives beyond their imagination. However, there is a push to ensure that artists remain central to this transformation, and that AI’s role in cinema respects creators’ rights and protections.

As AI technology evolves, Hollywood faces a critical choice: embrace it responsibly instead of the risk of being overtaken by powerful tech companies.

For more information on these topics, visit diplomacy.edu.

ChatGPT adds one million users per hour with AI image tool

OpenAI’s ChatGPT is experiencing unprecedented growth following the release of its new AI image generation feature. According to CEO Sam Altman, the AI tool has been attracting one million new users per hour, spurred by what he described as ‘biblical demand.’

The GPT-4o model, now available to free users, has quickly become a sensation, drawing comparisons to the app’s initial launch 26 months ago.

The latest update allows users to generate highly detailed visuals, including logos, diagrams, and imagery with precise meaning. OpenAI has touted GPT-4o’s ability to accurately render text, follow prompts closely, and draw from its extensive knowledge base to produce tailored visuals.

Popular uses include crafting images in the style of Studio Ghibli, which has led to a flood of creative outputs shared across social media platforms.

However, the overwhelming popularity has strained OpenAI’s resources. Sam Altman acknowledged the situation by noting that their GPUs were ‘melting’ under the pressure.

To manage the surge, OpenAI implemented temporary rate limits to enhance efficiency and maintain service quality.

The company’s success has also attracted significant investment, with OpenAI securing $40 billion in funding, bringing its valuation to $300 billion.

Since launching in November 2022, ChatGPT has amassed 500 million users, solidifying its position as one of the most influential AI platforms globally.

For more information on these topics, visit diplomacy.edu.

Agentic AI gains traction with Amazon’s Nova Act and OpenAI’s open-weight model

The competition to define the next era of agentic AI—systems capable of planning, reasoning, and executing tasks—continues to intensify as major tech firms introduce new AI models with varying levels of autonomy and openness.

Amazon has launched Nova Act, an AI-powered browser agent that operates much like a human user, handling low-risk, automated tasks. Meanwhile, OpenAI is preparing to release an open-weight language model, providing developers with access to its trained parameters for custom AI applications.

At the same time, China’s Zhipu AI has unveiled a free AI assistant aimed at strengthening its domestic market presence and competing with Western firms.

The launches signal a major shift in AI adoption strategies. While agentic AI holds vast potential, experts caution that its deployment in high-stakes enterprise environments requires further refinement.

‘The launch of Amazon Nova Act seems well suited for public-facing use cases with minimal risks, but enterprise adoption will take longer due to workflow complexity and operational concerns,’ said Sharath Srinivasamurthy, Associate VP of Research at IDC.

OpenAI’s move toward an open-weight model could accelerate AI customization efforts, aligning with growing enterprise demands for greater control.

Unlike fully open-source models that release training data and code, open-weight models provide access to trained parameters, enabling businesses to adapt AI to their specific needs.

‘The world is shifting towards open models,’ Srinivasamurthy added. ‘With Meta and DeepSeek gaining traction for their open approach, OpenAI’s move was only a matter of time. This will make the market even more competitive.’

Zhipu AI’s rise also introduces geopolitical factors, as China seeks to expand its AI leadership. The growing presence of Chinese AI firms may push multinational companies to reassess partnerships and compliance strategies in different markets.

‘The ascent of Zhipu AI highlights China’s increasing role in AI innovation, urging global businesses to carefully evaluate AI governance, security, and competitive strategy,’ said Abhivyakti Sengar, Practice Director at Everest Group.

As AI adoption accelerates, the evolving balance between control, performance, and openness is set to shape enterprise strategies in an increasingly competitive landscape.

For more information on these topics, visit diplomacy.edu.

Windows 11 update brings AI-powered Live Captions and image editing

Microsoft is making its AI-powered features more widely available across Copilot Plus PCs, extending support to Intel and AMD devices.

Among the most notable additions is Live Captions, which translates audio into English subtitles in real time from dozens of languages. Initially tested on these devices last December, Live Captions is now accessible through the latest Windows 11 update.

The update also introduces Cocreator, an AI-powered tool in Paint that generates images based on a user’s text description and drawings.

Additionally, Microsoft is expanding access to its AI image editor and generator in the Photos app, offering more creative possibilities to users.

Previously, these AI features were exclusive to Copilot Plus PCs with Qualcomm chips. While Microsoft has been testing Recall, an AI-driven screenshot search feature, on Intel and AMD devices, there’s no confirmed timeline for its full rollout.

Microsoft is also enhancing Voice Access, an accessibility feature that enables users to control their PC through voice commands. The update improves natural language understanding on Qualcomm-powered Copilot Plus PCs, with plans to bring this functionality to Intel and AMD devices in the future.

Additionally, Microsoft is expanding language translation, allowing users to translate 27 languages into Simplified Chinese.

As Microsoft continues integrating AI across Windows 11, these updates bring greater accessibility, creativity, and multilingual support to more users, solidifying AI’s role in everyday computing.

For more information on these topics, visit diplomacy.edu.

MetaAI launches in Europe amid data concerns

Meta has resumed the roll-out of its MetaAI across Europe after halting the launch last year due to regulatory uncertainty.

The Irish Data Protection Commission (DPC) still has questions regarding Meta’s AI tool, particularly in relation to its use of personal data from Facebook and Instagram users to train large language models.

The company has been in discussions with the DPC, but instead of an agreement, it remains under review as the tool continues to roll out.

MetaAI was first introduced in the US in September 2023, followed by India in June 2024, and the UK in October. It enables users to interact with a chat function across Facebook, Instagram, Messenger, and WhatsApp.

However, its expansion in Europe faced delays last summer due to concerns raised by the Irish privacy watchdog.

The company has expressed confidence in its compliance with the EU’s data protection laws and has been transparent with the DPC about its launch. However, failure to comply with the General Data Protection Regulation (GDPR) could lead to significant fines.

Additionally, certain aspects of MetaAI fall under the scope of Europe’s Digital Services Act (DSA), which requires the company to meet specific standards on user safety and transparency.

The European Commission has indicated it is waiting for a risk assessment from Meta to ensure that the tool complies with DSA obligations. While initial elements may not be directly relevant to the DSA, the Commission will continue to monitor the deployment closely.

For more information on these topics, visit diplomacy.edu.

Amazon unveils Nova Act to enhance AI capabilities

Amazon has launched Nova Act, a general-purpose AI agent capable of controlling web browsers to perform simple tasks. Along with the new agent, Amazon is releasing the Nova Act SDK, enabling developers to create agent prototypes.

The tool will also power key features of the upcoming Alexa+ upgrade, a generative AI-enhanced version of Amazon’s voice assistant.

Developed by Amazon’s AGI lab, Nova Act is designed to automate tasks such as ordering food or making reservations. Although the model is currently a research preview, Amazon claims Nova Act outperforms competitors like OpenAI’s Operator and Anthropic’s Computer Use in internal tests.

The toolkit, available on nova.amazon.com, allows developers to integrate AI agents into applications that can navigate websites, fill forms, and interact with digital content.

Despite its early stage, Nova Act is seen as a significant step in the development of superintelligent AI, with Amazon’s AGI lab aiming to make AI agents reliable and effective across various tasks.

Instead of AI agents from other companies that have faced challenges like slow response times and error-prone performance, Amazon hopes that Nova Act will address these issues, potentially providing a competitive edge in the AI market.

The success of Nova Act could also play a crucial role in the success of Alexa+ and Amazon’s broader AI strategy.

For more information on these topics, visit diplomacy.edu.

Runway expands AI video capabilities with Gen-4

Runway has unveiled Gen-4, its most advanced AI-powered video generator yet, promising superior character consistency, realistic motion, and world understanding.

The model is now available to individual and enterprise users, allowing them to generate dynamic videos using visual references and text-based instructions.

Backed by investors such as Google and Nvidia, Runway faces fierce competition from OpenAI and Google in the AI video space. The company has differentiated itself by securing Hollywood partnerships and investing heavily in AI-generated filmmaking.

However, it remains tight-lipped about its training data, raising concerns over copyright issues.

Runway is currently embroiled in a lawsuit from artists accusing the company of training its models on copyrighted works instead of getting permission. The company claims fair use as a defence.

Meanwhile, it is reportedly seeking new funding at a $4 billion valuation, with hopes of reaching $300 million in annual revenue. As AI video tools advance, concerns grow over their impact on jobs in the entertainment industry, with thousands of positions at risk.

For more information on these topics, visit diplomacy.edu.

Apple expands AI features with new update

Apple Intelligence is expanding with new features, including Priority Notifications, which highlight time-sensitive alerts for users. This update is part of iOS 18.4, iPadOS 18.4, and macOS Sequoia 15.4, rolling out globally.

The AI suite is now available in more languages and has launched in the EU for iPhone and iPad users.

Additional improvements include a new Sketch style in Image Playground and the ability to generate ‘memory movies’ on Mac using simple text descriptions. Vision Pro users in the US can now access Apple Intelligence features like Writing Tools and Genmoji.

Apple’s AI rollout has been gradual since its introduction at WWDC last year, with features arriving in stages.

The update also brings fresh emojis, child safety enhancements, and the debut of Apple News+ Food, further expanding Apple’s digital ecosystem.

For more information on these topics, visit diplomacy.edu.

OpenAI expands image generator access to all users

OpenAI has made its image generator, powered by the GPT-4o model, accessible to all users, CEO Sam Altman announced on X. Previously, this feature was available only to paying ChatGPT subscribers.

While there is no clear indication of how many images free-tier users can create, Altman previously mentioned a possible limit of three per day.

The tool has seen massive demand since its launch, with Altman joking that OpenAI’s GPUs were ‘melting’ under the pressure. However, it has also sparked controversy, particularly after users began generating images in the style of Studio Ghibli, raising copyright concerns.

Others have used the generator to create fake receipts, such as restaurant bills. OpenAI has responded by stating that all AI-generated images contain metadata identifying them and that the company takes action when violations occur.

In a major financial development, OpenAI has secured $40 billion in funding from SoftBank, valuing the company at $300 billion. The company also revealed that ChatGPT now boasts 500 million weekly active users and 700 million monthly active users, marking a significant milestone in its growth.

For more information on these topics, visit diplomacy.edu.