Content Policy

Inside OpenAI’s battle to protect AI from prompt injection attacks

OpenAI has identified prompt injection as one of the most pressing new challenges in AI security. As AI systems gain the ability to browse the web, handle personal data and act on users’ behalf, they become targets for malicious instructions hidden within online content.

These attacks, known as prompt injections, can trick AI models into taking unintended actions or revealing sensitive information.

To counter the issue, OpenAI has adopted a multi-layered defence strategy that combines safety training, automated monitoring and system-level security protections. The company’s research into ‘Instruction Hierarchy’ aims to help models distinguish between trusted and untrusted commands.

Continuous red-teaming and automated detection systems further strengthen resilience against evolving threats.

OpenAI also provides users with greater control, featuring built-in safeguards such as approval prompts before sensitive actions, sandboxing for code execution, and ‘Watch Mode’ when operating on financial or confidential sites.

These measures ensure that users remain aware of what actions AI agents perform on their behalf.

While prompt injection remains a developing risk, OpenAI expects adversaries to devote significant resources to exploiting it. The company continues to invest in research and transparency, aiming to make AI systems as secure and trustworthy as a cautious, well-informed human colleague.

Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot!

Cars.com launches Carson AI to transform online car shopping

The US tech company, Cars.com, has unveiled Carson, a multilingual AI search engine designed to revolutionise the online car shopping experience.

Instead of relying on complex filters, Carson interprets natural language queries such as ‘a reliable car for a family of five’ or ‘a used truck under $30,000’, instantly producing targeted results tailored to each shopper’s needs.

A new AI feature that already powers around 15% of all web and mobile searches on Cars.com, with early data showing that users engaging with Carson return to the site twice as often and save three times more vehicles.

They also generate twice as many leads and convert 30% more frequently from search to vehicle detail pages.

Cars.com aims to simplify decision-making for its 25 million monthly shoppers, 70% of whom begin their search without knowing which brand or model to choose.

Carson helps these undecided users explore lifestyle, emotional and practical preferences while guiding them through Cars.com’s award-winning listings.

Further updates will introduce AI-generated summaries, personalised comparisons and search refinement suggestions.

Cars.com’s parent company, Cars Commerce, plans to expand its use of AI-driven tools to strengthen its role at the forefront of automotive retail innovation, offering a more efficient and intelligent marketplace for both consumers and dealerships.

Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot!

Appfigures revises iOS estimates as Sora’s launch on Android launch leaps ahead

Sora’s Android launch outpaced its iOS debut, garnering an estimated 470,000 first-day installs across seven markets, according to Appfigures. Broader regional availability, plus the end of invite-only access in top markets, boosted uptake.

OpenAI’s iOS rollout was limited to the US and Canada via invitations, which capped early growth despite strong momentum. The iOS app nevertheless surpassed one million installs in its first week and still ranks highly in the US App Store’s Top Free chart.

Revised Appfigures modelling puts day-one iOS installs at ~110,000 (up from 56,000), with ~69,300 from the US. On Android, availability spans the US, Canada, Japan, South Korea, Taiwan, Thailand, and Vietnam. First-day US installs were ~296,000, showing sustained demand beyond the iOS launch.

Sora allows users to generate videos from text prompts and animate themselves or friends via ‘Cameos’, sharing the results in a TikTok-style vertical feed. Engagement features for creation and discovery are driving word of mouth and repeat use across both platforms.

Competition in mobile AI video and assistants is intensifying, with Meta AI expanding its app in Europe on the same day. Market share will hinge on geographic reach, feature velocity, creator tools, and distribution via app store charts and social feeds.

Would you like to learn more about AI, tech, and digital diplomacy? If so, ask our Diplo chatbot!

Amazon introduces Kindle Translate for global eBook reach

Amazon has introduced Kindle Translate, an AI-powered translation service designed to help Kindle Direct Publishing (KDP) authors reach readers worldwide.

The beta service currently supports translations between English and Spanish, and from German to English, enabling authors to make their eBooks accessible in multiple languages.

Authors can manage translations, set prices, and publish fully formatted books within days, with automated accuracy checks ensuring seamless publication.

The service addresses the limited availability of multilingual titles on Amazon, where fewer than 5% of books are offered in multiple languages. Kindle Translate helps authors reach a wider audience and boost earnings with a cost-effective and reliable translation solution.

Early adopters, including independent authors Roxanne St Claire and Kristen Painter, praised the service for enabling wider international readership and boosting book revenue.

Translated eBooks will be available for purchase on Amazon, with clear labels and preview samples for readers. Titles are eligible for KDP Select and Kindle Unlimited, with more languages planned to expand the library of translated works.

Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot! 

OpenAI unveils Teen Safety Blueprint for responsible AI

OpenAI has launched the Teen Safety Blueprint to guide responsible AI use for young people. The roadmap guides policymakers and developers on age-appropriate design, safeguards, and research to protect teen well-being and promote opportunities.

The company is implementing these principles across its products without waiting for formal regulation. Recent measures include stronger safeguards, parental controls, and an age-prediction system to customise AI experiences for under-18 users.

OpenAI emphasises that protecting teens is an ongoing effort. Collaboration with parents, experts, and young people will help improve AI safety continuously while shaping how technology can support teens responsibly over the long term.

Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot! 

Coca-Cola enhances its AI-powered Christmas ad to fix last year’s visual flaws

Coca-Cola has released an improved AI-generated Christmas commercial after last year’s debut campaign drew criticism for its unsettling visuals.

The latest ‘Holidays Are Coming’ ads, developed in part by San Francisco-based Silverside, showcase more natural animation and a wider range of festive creatures, instead of the overly lifelike characters that previously unsettled audiences.

The new version avoids the ‘uncanny valley’ effect that plagued 2024’s ads. The use of generative AI by Coca-Cola reflects a wider advertising trend focused on speed and cost efficiency, even as creative professionals warn about its potential impact on traditional jobs.

Despite the efficiency gains, AI-assisted advertising remains labour-intensive. Teams of digital artists refine the content frame by frame to ensure realistic and emotionally engaging visuals.

Industry data show that 30% of commercials and online videos in 2025 were created or enhanced using generative AI, compared with 22% in 2023.

Coca-Cola’s move follows similar initiatives by major firms, including Google’s first fully AI-generated ad spot launched last month, signalling that generative AI is now becoming a mainstream creative tool across global marketing.

Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot!

How Google uses AI to support teachers and inspire students

Google is redefining education with AI designed to enhance learning, rather than replace teachers. The company has unveiled new tools grounded in learning science to support both educators and students, aiming to make learning more effective, efficient and engaging.

Through its Gemini platform, users can follow guided learning paths that encourage discovery rather than passive answers.

YouTube and Search now include conversational features that allow students to ask questions as they learn, while NotebookLM can transform personal materials into quizzes or immersive study aids.

Instructors can also utilise Google Classroom’s free AI tools for lesson planning and administrative support, thereby freeing up time for direct student engagement.

Google emphasises that its goal is to preserve the human essence of education while using AI to expand understanding. The company also addresses challenges linked to AI in learning, such as cheating, fairness, accuracy and critical thinking.

It is exploring assessment models that cannot be easily replicated by AI, including debates, projects, and oral examinations.

The firm pledges to develop its tools responsibly by collaborating with educators, parents and policymakers. By combining the art of teaching with the science of AI-driven learning, Google seeks to make education more personal, equitable and inspiring for all.

Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot!

Meta rejects French ruling over gender bias in Facebook job ads

Meta has rejected a decision by France’s Défenseur des Droits that found its Facebook algorithm discriminates against users based on gender in job advertising. The case was brought by Global Witness and women’s rights groups Fondation des Femmes and Femmes Ingénieures, who argued that Meta’s ad system violates French anti-discrimination law.

The regulator ruled that Facebook’s system treats users differently according to gender when displaying job opportunities, amounting to indirect discrimination. It recommended Meta Ireland and Facebook France make adjustments within three months to prevent gender-based bias.

A Meta spokesperson said the company disagrees with the finding and is ‘assessing its options.’ The complainants welcomed the decision, saying it confirms that platforms are not exempt from laws prohibiting gender-based distinctions in recruitment advertising.

Lawyer Josephine Shefet, representing the groups, said the ruling marks a key precedent. ‘The decision sends a strong message to all digital platforms: they will be held accountable for such bias,’ she said.

Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot!

OpenAI outlines roadmap for AI safety, accountability and global cooperation

New recommendations have been published by OpenAI for managing rapid advances in AI, stressing the need for shared safety standards, public accountability, and resilience frameworks.

The company warned that while AI systems are increasingly capable of solving complex problems and accelerating discovery, they also pose significant risks that must be addressed collaboratively.

According to OpenAI, the next few years could bring systems capable of discoveries once thought centuries away.

The firm expects AI to transform health, materials science, drug development and education, while acknowledging that economic transitions may be disruptive and could require a rethinking of social contracts.

To ensure safe development, OpenAI proposed shared safety principles among frontier labs, new public oversight mechanisms proportional to AI capabilities, and the creation of a resilience ecosystem similar to cybersecurity.

It also called for regular reporting on AI’s societal impact to guide evidence-based policymaking.

OpenAI reiterated that the goal should be to empower individuals by making advanced AI broadly accessible, within limits defined by society, and to treat access to AI as a foundational public utility in the years ahead.

Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot!

Snap brings Perplexity’s answer engine into Chat for nearly a billion users

Starting in early 2026, Perplexity’s AI will be integrated into Snapchat’s Chat, accessible to nearly 1 billion users. Snapchatters can ask questions and receive concise, cited answers in-app. Snap says the move reinforces its position as a trusted, mobile-first AI platform.

Under the deal, Perplexity will pay Snap $400 million in cash and equity over a one-year period, tied to the global rollout. Revenue contribution is expected to begin in 2026. Snap points to its 943 million MAUs and reaches over 75% of 13–34-year-olds in 25+ countries.

Perplexity frames the move as meeting curiosity where it occurs, within everyday conversations. Evan Spiegel says Snap aims to make AI more personal, social, and fun, woven into friendships and conversations. Both firms pitch the partnership as enhancing discovery and learning on Snapchat.

Perplexity joins, rather than replaces, Snapchat’s existing My AI. Messages sent to Perplexity will inform personalisation on Snapchat, similar to My AI’s current behaviour. Snap claims the approach is privacy-safe and designed to provide credible, real-time answers from verifiable sources.

Snap casts this as a first step toward a broader AI partner platform inside Snapchat. The companies plan creative, trusted ways for leading AI providers to reach Snap’s global community. The integration aims to enable seamless, in-chat exploration while keeping users within Snapchat’s product experience.

Would you like to learn more about AI, tech, and digital diplomacy? If so, ask our Diplo chatbot!