Olga Loiek, a 21-year-old University of Pennsylvania student from Ukraine, experienced a disturbing twist after launching her YouTube channel last November. Her image was hijacked and manipulated through AI to create digital alter egos on Chinese social media platforms. These AI-generated avatars, such as ‘Natasha,’ posed as Russian women fluent in Chinese, promoting pro-Russian sentiments and selling products like Russian candies. These fake accounts amassed hundreds of thousands of followers in China, far surpassing Loiek’s own online presence.
Loiek’s experience highlights a broader trend of AI-generated personas on Chinese social media, presenting themselves as supportive of Russia and fluent in Chinese while selling various products. Experts reveal that these avatars often use clips of real women without their knowledge, aiming to appeal to single Chinese men. Some posts include disclaimers about AI involvement, but the followers and sales figures remain significant.
Why does it matter?
These events underscore the ethical and legal concerns surrounding AI’s misuse. As generative AI systems like ChatGPT become more widespread, issues related to misinformation, fake news, and copyright violations are growing.
In response, governments are starting to regulate the industry. China proposed guidelines to standardise AI by 2026, while the EU’s new AI Act imposes strict transparency requirements. However, experts like Xin Dai from Peking University warn that regulations struggle to keep pace with rapid AI advancements, raising concerns about the unchecked proliferation of AI-generated content worldwide.
Anthropic, a startup backed by Google and Amazon, has introduced a new AI model named Claude 3.5 Sonnet alongside a revamped user interface to enhance productivity. The release comes just three months after the launch of its Claude 3 family of AI models. Claude 3.5 Sonnet surpasses its predecessor, Claude 3 Opus, in benchmark exam performance, speed, and cost efficiency, being five times cheaper for developers.
CEO Dario Amodei emphasised AI’s flexibility and rapid advancement, noting that, unlike physical products, AI models can be quickly updated and improved. Anthropic’s latest technology is now available for free on Claude.ai and through an iOS app. Additionally, users can opt into a new feature called ‘Artifacts,’ which organises generated content in a side window, facilitating collaborative work and the production of finished products.
Anthropic’s rapid development cycle reflects the competitive nature of the AI industry, with companies like OpenAI and Google also pushing forward with significant AI advancements. The startup plans to release more models, including Claude 3.5 Opus, within the year while focusing on safety and usability.
The US House of Representatives is unlikely to pass broad AI regulation this year. House Majority Leader Steve Scalise said that he opposes extensive regulations, fearing they might hinder the US in AI development compared to China. Instead, he suggests focusing on existing laws and targeted fixes rather than creating new regulatory structures.
This stance contrasts with Senate Majority Leader Chuck Schumer’s proposal, whose bipartisan AI working group report recommended a $32 billion annual investment in non-defense AI innovation and a comprehensive regulatory framework. The House’s bipartisan AI task force is also cautious about large-scale regulations.
Chair Rep. Jay Obernolte suggests that some targeted AI legislation might be feasible, while Rep. French Hill advocates for a sector-specific review under existing laws rather than a broad regulatory framework. This division between the House and Senate reduces the likelihood of significant AI legislation this year, but the House may consider smaller, urgent AI-related bills to address immediate issues.
Why does it matter?
The US Congress has seen a surge in AI legislation from both the Senate and House, by the rise of advanced AI models like ChatGPT and DeepAI, and growing issues with ‘deepfake’ content, particularly around elections and scams. However, this division reduces the likelihood of significant AI legislation this year, though smaller, urgent AI-related bills may still be approved.
Amazon has expanded its generative AI tools for product listings to sellers in France, Germany, Italy, Spain, and the UK. These tools, designed to streamline the process of creating and enhancing product listings, can generate product descriptions, titles, and details and fill in any missing information. The rollout follows an initial introduction in the US and a quieter launch in the UK earlier this month.
The new AI tools aim to help sellers list products more quickly by allowing them to enter relevant keywords or upload product photos, after which the AI generates a product title, bullet points, and descriptions. While the AI-generated content can be edited, Amazon advises sellers to review the generated listings thoroughly to avoid inaccuracies. The company continuously improves these tools to make them more effective and user-friendly.
Earlier this year, Amazon also introduced a tool enabling sellers to create product listings by posting a URL to their existing website, though it remains uncertain when this feature will be available outside the US. The expansion of AI tools to European markets raises regulatory concerns, particularly regarding GDPR and the Digital Services Act, which require transparency in AI applications.
Why does it matter?
Despite these regulatory challenges, Amazon’s use of generative AI marks a significant advancement in e-commerce. By leveraging diverse sources of information, Amazon’s AI models can infer product details with high accuracy, improving the quality and efficiency of product listings at scale. However, the precise data used to train these models remains unclear, highlighting ongoing concerns about data privacy and usage.
In a post on his platform X, tech boss Elon Musk revealed his partnership with Dell Technologies and Super Micro Computer to provide server racks set to power Musk’s AI startup, xAI. According to Dell CEO Michael Dell, the company is responsible for manufacturing half the microchips for the project. To this end, Dell has partnered with Nvidia Corporation, a multinational technology company that supplies electronic chips for computer motherboard chipsets, smartphones and game consoles. The duo, together with Super Micro Computers, will build xAI’s supercomputer, which will, in turn, power the startup’s chatbot, Grok.
Musk’s xAI, a rival to Microsoft-backed OpenAI and Alphabet’s Google, opened its doors last year, and its chatbot is already on version 2. Musk’s timeline for Grok 3 is fall 2025. The chatbot requires 100,000 Nvidia H100 chips to run. Grok 2 utilised 20,000 Nvidia H100 graphic processing units. The Nvidia H100 GPU chip is the most powerful GPU chip on the market and is designed for AI applications only. The chips are in short supply, and on average, one costs $25,000.
A report published by venture capital firm Accel shows the state of affairs of Europe and Israel’s generative AI (GenAI). That type of AI is able to generate text, images, sounds, videos and other mediums based on datasets upon which it is trained. Four countries in the region are leading the pack, with the UK and France coming up as winners in their distinct categories.
The UK leads with 30% of the 221 GenAI startups analysed, followed by Germany (14%), Israel (13%) and France (11%). The UK’s strong lead can be explained by a long tradition of tech and AI development over the last quarter century. UK universities have partnered with tech giants to create AI research hubs throughout the country, setting the perfect grounds for attracting AI talent. UK based AI giants like DeepMind, Microsoft and Meta, have seen many of their former employees go on to create their own GenAI startups. 12% of all GenAI startups surveyed in the UK had at least one founder who had worked at the British AI pioneer DeepMind, the report said.
Despite coming in fourth place by quantity of GenAI startups, France comes in first in terms of funding at $2.29 billion. Next, at half of that, is the UK with $1.15 billion, Israel at $1.04 billion and Germany at $636 million. The funding boom is both a symptom and a cause of the country becoming a new AI hub for Europe. French universities such as École Polytechnique, Université Pierre et Marie Curie and École normale Supérieure have educated and given work experience to the many maths and engineering students starting GenAI startups. In return, this has led Google, Meta and Kyutai to open facilities in Paris to attract that potential.
Three of the region’s best funded companies are French, with Mistral, Europe’s competitor to OpenAI, raising over €600 million alone. Cooperation between universities and tech giants in Israel and Germany are also a main source of growth for their respective AI sectors.
In a bold move highlighting the intersection of technology and politics, businessman Steve Endacott is running in the 4 July national election in Britain, aiming to become a member of parliament (MP) with the aid of an AI-generated avatar. The campaign leaflet for Endacott features not his own face but that of an AI avatar dubbed ‘AI Steve.’ The initiative, if successful, would result in the world’s first AI-assisted lawmaker.
Endacott, founder of Neural Voice, presented his AI avatar to the public in Brighton, engaging with locals on various issues through real-time interactions. The AI discusses topics like LGBTQ rights, housing, and immigration and then offers policy ideas, seeking feedback from citizens. Endacott aims to demonstrate how AI can enhance voter access to their representatives, advocating for a reformed democratic process where people are more connected to their MPs.
Despite some scepticism, with concerns about the effectiveness and trustworthiness of an AI MP, Endacott insists that the AI will serve as a co-pilot, formulating policies reviewed by a group of validators to ensure security and integrity. The Electoral Commission clarified that the elected candidate would remain the official MP, not the AI. While public opinion is mixed, the campaign underscores the growing role of AI in various sectors and sparks an important conversation about its potential in politics.
Founder Masayoshi Son announced that Japan’s SoftBank Group plans to expand its power generation business in the US to support global generative AI projects. SB Energy, backed by SoftBank, focuses on developing and operating renewable energy projects across the US. The initiative aligns with SoftBank’s strategy to explore new investment opportunities outside Japan.
Why does it matter?
At the annual shareholder meeting of SoftBank Corp, the group’s telecom arm, Son highlighted the importance of seeking innovative investments. He emphasised that SoftBank’s future growth would rely on identifying and nurturing emerging technologies and markets beyond Japan.
The current strategy reflects SoftBank’s commitment to advancing its global presence and influence in the tech and renewable energy sectors.
Ilya Sutskever, co-founder and former chief scientist at OpenAI, announced on Wednesday the launch of a new AI company named Safe Superintelligence. The company aims to create a secure AI environment amidst the competitive generative AI industry. Based in Palo Alto and Tel Aviv, Safe Superintelligence aims to prioritise safety and security over short-term commercial pressures.
Sutskever made the announcement on social media, emphasising the company’s focused approach without the distractions of traditional management overhead or product cycles. Joining him as co-founders are Daniel Levy, a former OpenAI researcher, and Daniel Gross, co-founder of Cue and former AI lead at Apple.
Sutskever’s departure from Microsoft-backed OpenAI in May followed his involvement in the dramatic firing and rehiring of CEO Sam Altman in November of the previous year. His new venture underscores a commitment to advancing AI technology in a manner that ensures safety and long-term progress.
China’s AI military commander substitutes for human military leaders in simulated war games hosted by the Joint Operations College of the National Defence University, amidst growing tensions with the US over the use of militarised AI in combat. The bots, the first of their kind, are completely automated, possess the perception and reasoning skills of human military leaders, and are learning at an exponential rate. They have also been programmed to illustrate the weaknesses of some of the country’s most celebrated military leaders such as General Peng Dehuai, and General Lin Biao.
The AI arms race between the two countries can be likened to the chicken and egg analogy, in that both countries have expressed interest in regulating the use of these unmanned implements on the battlefield; yet, there are increasing media coverage of either on-going experiments or caged prototypes in both countries. These include the rifle-toting robot dogs, and surveillance and attack drones, some of which reportedly have already been used in battlefields in Gaza and in the Ukraine. The situation renders international rule-making in the space increasingly difficult, particularly as other players, such as NATO seek to ramp up investments in tech-driven defence systems.
