Consumer protection

Google removes Gemma AI model following defamation claims

Google has removed its Gemma AI model from AI Studio after US Senator Marsha Blackburn accused it of producing false sexual misconduct claims about her. The senator said Gemma fabricated an incident allegedly from her 1987 campaign, citing nonexistent news links to support the claim.

Blackburn described the AI’s response as defamatory and demanded action from Google.

The controversy follows a similar case involving conservative activist Robby Starbuck, who claims Google’s AI tools made false accusations about him. Google acknowledged that AI’ hallucinations’ are a known issue but insisted it is working to mitigate such errors.

Blackburn argued these fabrications go beyond harmless mistakes and represent real defamation from a company-owned AI model.

Google stated that Gemma was never intended as a consumer-facing tool, noting that some non-developers misused it to ask factual questions. The company confirmed it would remove the model from AI Studio while keeping it accessible via API for developers.

The incident has reignited debates over AI bias and accountability. Blackburn highlighted what she sees as a consistent pattern of conservative figures being targeted by AI systems, amid wider political scrutiny over misinformation and AI regulation.

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

When AI LLMs ‘think’ more, groups suffer, CMU study finds

Researchers at Carnegie Mellon University report that stronger-reasoning language models (LLMs) act more selfishly in groups, reducing cooperation and nudging peers toward self-interest. Concerns grow as people ask AI for social advice.

In a Public Goods test, non-reasoning models shared 96 percent; a reasoning model shared 20 percent. Adding a few reasoning steps cut cooperation nearly in half. Reflection prompts also reduced sharing.

Mixed groups showed spillover. Reasoning agents dragged down collective performance by 81 percent, spreading self-interest. Users may over-trust ‘rational’ advice that justifies uncooperative choices at work or in class.

Comparisons spanned LLMs from OpenAI, Google, DeepSeek, and Anthropic. Findings point to the need to balance raw reasoning with social intelligence. Designers should reward cooperation, not only optimise individual gain.

The paper ‘Spontaneous Giving and Calculated Greed in Language Models’ will be presented at EMNLP 2025, with a preprint on arXiv. Authors caution that more intelligent AI is not automatically better for society.

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

Australian police create AI tool to decode predators’ slang

Australian police are developing an AI tool with Microsoft to decode slang and emojis used by online predators. The technology is designed to interpret coded messages in digital conversations to help investigators detect harmful intent more quickly.

Federal Police Commissioner Krissy Barrett said social media has become a breeding ground for exploitation, bullying, and radicalisation. The AI based prototype, she explained, could allow officers to identify threats earlier and rescue children before abuse occurs.

Barrett also warned about the rise of so-called ‘crimefluencers’, offenders using social media trends to lure young victims, many of whom are pre-teen or teenage girls. Australian authorities believe understanding modern online language is key to disrupting their methods.

The initiative follows Australia’s new under-16 social media ban, due to take effect in December. Regulators worldwide are monitoring the country’s approach as governments struggle to balance online safety with privacy and digital rights.

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

WhatsApp adds passkey encryption for safer chat backups

Meta is rolling out a new security feature for WhatsApp that allows users to encrypt their chat backups using passkeys instead of passwords or lengthy encryption codes.

A feature for WhatsApp that enables users to protect their backups with biometric authentication such as fingerprints, facial recognition or screen lock codes.

WhatsApp became the first messaging service to introduce end-to-end encrypted backups over four years ago, and Meta says the new update builds on that foundation to make privacy simpler and more accessible.

With passkey encryption, users can secure and access their chat history easily without the need to remember complex keys.

The feature will be gradually introduced worldwide over the coming months. Users can activate it by going to WhatsApp settings, selecting Chats, then Chat backup, and enabling end-to-end encrypted backup.

Meta says the goal is to make secure communication effortless while ensuring that private messages remain protected from unauthorised access.

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

A licensed AI music platform emerges from UMG and Udio

UMG and Udio have struck an industry-first deal to license AI music, settle litigation, and launch a 2026 platform that blends creation, streaming, and sharing in a licensed environment. Training uses authorised catalogues, with fingerprinting, filtering, and revenue sharing for artists and songwriters.

Udio’s current app stays online during the transition under a walled garden, with fingerprinting, filtering, and other controls added ahead of relaunch. Rights management sits at the core: licensed inputs, transparent outputs, and enforcement that aims to deter impersonation and unlicensed derivatives.

Leaders frame the pact as a template for a healthier AI music economy that aligns rightsholders, developers, and fans. Udio calls it a way to champion artists while expanding fan creativity, and UMG casts it as part of its broader AI partnerships across platforms.

Commercial focus extends beyond headline licensing to business model design, subscriptions, and collaboration tools for creators. Expect guardrails around style guidance, attribution, and monetisation, plus pathways for official stems and remix packs so fan edits can be cleared and paid.

Governance will matter as usage scales, with audits of model inputs, takedown routes, and payout rules under scrutiny. Success will be judged on artist adoption, catalogue protection, and whether fans get safer ways to customise music without sacrificing rights.

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

OpenAI unveils new gpt-oss-safeguard models for adaptive content safety

Yesterday, OpenAI launched gpt-oss-safeguard, a pair of open-weight reasoning models designed to classify content according to developer-specified safety policies.

Available in 120b and 20b sizes, these models allow developers to apply and revise policies during inference instead of relying on pre-trained classifiers.

They produce explanations of their reasoning, making policy enforcement transparent and adaptable. The models are downloadable under an Apache 2.0 licence, encouraging experimentation and modification.

The system excels in situations where potential risks evolve quickly, data is limited, or nuanced judgements are required.

Unlike traditional classifiers that infer policies from pre-labelled data, gpt-oss-safeguard interprets developer-provided policies directly, enabling more precise and flexible moderation.

The models have been tested internally and externally, showing competitive performance against OpenAI’s own Safety Reasoner and prior reasoning models. They can also support non-safety tasks, such as custom content labelling, depending on the developer’s goals.

OpenAI developed these models alongside ROOST and other partners, building a community to improve open safety tools collaboratively.

While gpt-oss-safeguard is computationally intensive and may not always surpass classifiers trained on extensive datasets, it offers a dynamic approach to content moderation and risk assessment.

Developers can integrate the models into their systems to classify messages, reviews, or chat content with transparent reasoning instead of static rule sets.

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

US Internet Bill of Rights unveiled as response to global safety laws

A proposed US Internet Bill of Rights aims to protect digital freedoms as governments expand online censorship laws. The framework, developed by privacy advocates, calls for stronger guarantees of free expression, privacy, and access to information in the digital era.

Supporters argue that recent legislation such as the UK’s Online Safety Act, the EU’s Digital Services Act, and US proposals like KOSA and the STOP HATE Act have eroded civil liberties. They claim these measures empower governments and private firms to control online speech under the guise of safety.

The proposed US bill sets out rights including privacy in digital communications, platform transparency, protection against government surveillance, and fair access to the internet. It also calls for judicial oversight of censorship requests, open algorithms, and the protection of anonymous speech.

Advocates say the framework would enshrine digital freedoms through federal law or constitutional amendment, ensuring equal access and privacy worldwide. They argue that safeguarding free and open internet access is vital to preserve democracy and innovation.

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

Character.ai restricts teen chat access on its platform

The AI chatbot service, Character.ai, has announced that teenagers can no longer chat with its AI characters from 25 November.

Under-18s will instead be limited to generating content such as videos, as the platform responds to concerns over risky interactions and lawsuits in the US.

Character.ai has faced criticism after avatars related to sensitive cases were discovered on the site, prompting safety experts and parents to call for stricter measures.

The company cited feedback from regulators and safety specialists, explaining that AI chatbots can pose emotional risks for young users by feigning empathy or providing misleading encouragement.

Character.ai also plans to introduce new age verification systems and fund a research lab focused on AI safety, alongside enhancing role-play and storytelling features that are less likely to place teens in vulnerable situations.

Safety campaigners welcomed the decision but emphasised that preventative measures should have been implemented.

Experts warn the move reflects a broader shift in the AI industry, where platforms increasingly recognise the importance of child protection in a landscape transitioning from permissionless innovation to more regulated oversight.

Analysts note the challenge for Character.ai will be maintaining teen engagement without encouraging unsafe interactions.

Separating creative play from emotionally sensitive exchanges is key, and the company’s new approach may signal a maturing phase in AI development, where responsible innovation prioritises the protection of young users.

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

Microsoft faces Australian lawsuit over hidden AI subscription option

In a legal move that underscores growing scrutiny of digital platforms, the Australian Competition and Consumer Commission (ACCC) has filed a lawsuit in the Federal Court against Microsoft Corporation, accusing the company of misleading approximately 2.7 million Australian personal and family-plan subscribers of its Microsoft 365 service after integrating its AI assistant Copilot.

According to the ACCC, Microsoft raised subscription prices by 45 % for the Personal plan and 29 % for the Family plan after bundling Copilot starting 31 October 2024.

The regulator says Microsoft told consumers their only options were to pay the higher price with AI or cancel their subscription, while failing to clearly disclose a cheaper ‘Classic’ version of the plan without Copilot that remained available.

The ACCC argues Microsoft’s communications omitted the existence of that lower-priced plan unless consumers initiated the cancellation process. Chair Gina Cass-Gottlieb described this omission as ‘very serious conduct’ that deprived customers of informed choice.

The regulator is seeking penalties, consumer redress, injunctions and costs, with potential sanctions of AUS $50 million (or more) per breach.

This action signals a broader regulatory push into how major technology firms bundle AI features, raise prices and present options to consumers, an issue that ties into digital economy governance, consumer trust and platform accountability.

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

Wikipedia founder questions Musk’s Grokipedia accuracy

Speaking at the CNBC Technology Executive Council Summit in New York, Wikipedia founder Jimmy Wales has expressed scepticism about Elon Musk’s new AI-powered Grokipedia, suggesting that large language models cannot reliably produce accurate wiki entries.

Wales highlighted the difficulties of verifying sources and warned that AI tools can produce plausible but incorrect information, citing examples where chatbots fabricated citations and personal details.

He rejected Musk’s claims of liberal bias on Wikipedia, noting that the site prioritises reputable sources over fringe opinions. Wales emphasised that focusing on mainstream publications does not constitute political bias but preserves trust and reliability for the platform’s vast global audience.

Despite his concerns, Wales acknowledged that AI could have limited utility for Wikipedia in uncovering information within existing sources.

However, he stressed that substantial costs and potential errors prevent the site from entirely relying on generative AI, preferring careful testing before integrating new technologies.

Wales concluded that while AI may mislead the public with fake or plausible content, the Wiki community’s decades of expertise in evaluating information help safeguard accuracy. He urged continued vigilance and careful source evaluation as misinformation risks grow alongside AI capabilities.

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