OpenAI

EU considers classifying ChatGPT as a search engine under the DSA. What are the implications?

The European Commission is pondering whether OpenAI’s ChatGPT should be designated as a ‘Very Large Online Search Engine’ (VLOSE) under the Digital Services Act (DSA), a move that could reshape how generative AI tools are regulated across Europe.

OpenAI recently reported that ChatGPT’s search feature reached 120.4 million monthly users in the EU over the past six months, well above the 45 million threshold that triggers stricter obligations for major online platforms and search engines. The Commission confirmed it is reviewing the figures and assessing whether ChatGPT meets the criteria for designation.

The key question is whether ChatGPT’s live search function should be treated as an independent service or as part of the chatbot as a whole. Legal experts note that the DSA applies to intermediary services such as hosting platforms or search engines, categories that do not neatly encompass generative AI systems.

Implications for OpenAI

If designated, ChatGPT would be the first AI chatbot formally subject to DSA obligations, including systemic risk assessments, transparency reporting, and independent audits. OpenAI would need to evaluate how ChatGPT affects fundamental rights, democratic processes, and mental health, updating its systems and features based on identified risks.

‘As part of mitigation measures, OpenAI may need to adapt ChatGPT’s design, features, and functionality,’ said Laureline Lemoine of AWO. ‘Compliance could also slow the rollout of new tools in Europe if risk assessments aren’t planned in advance.’

The company could also face new data-sharing obligations under Article 40 of the DSA, allowing vetted researchers to request information about systemic risks and mitigation efforts, potentially extending to model data or training processes.

A test case for AI oversight

Legal scholars say the decision could set a precedent for generative AI regulation across the EU. ‘Classifying ChatGPT as a VLOSE will expand scrutiny beyond what’s currently covered under the AI Act,’ said Natali Helberger, professor of information law at the University of Amsterdam.

Experts warn the DSA would shift OpenAI from voluntary AI-safety frameworks and self-defined benchmarks to binding obligations, moving beyond narrow ‘bias tests’ to audited systemic-risk assessments, transparency and mitigation duties. ‘The DSA’s due diligence regime will be a tough reality check,’ said Mathias Vermeulen, public policy director at AWO.

Would you like to learn more aboutAI, 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!

ChatGPT offers wellness checks for long chat sessions

OpenAI has introduced new features in ChatGPT to encourage healthier use for people who spend extended periods chatting with the AI. Users may see a pop-up message reading ‘Just checking in. You’ve been chatting for a while, is this a good time for a break?’.

Users can dismiss it or continue, helping to prevent excessive screen time while staying flexible. The update also guides high-stakes personal decisions.

ChatGPT will not give direct advice on sensitive topics such as relationships, but instead asks questions and encourages reflection, helping users consider their options safely.

OpenAI acknowledged that AI can feel especially personal for vulnerable individuals. Earlier versions sometimes struggled to recognise signs of emotional dependency or distress.

The company is improving the model to detect these cases and direct users to evidence-based resources when needed, making long interactions safer and more mindful.

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

OpenAI Foundation to fund global health and AI safety projects

OpenAI has finalised its recapitalisation, simplifying its structure while preserving its core mission. The new OpenAI Foundation controls OpenAI Group PBC and holds about $130 billion in equity, making it one of history’s best-funded philanthropies.

The Foundation will receive further ownership as OpenAI’s valuation grows, ensuring its financial resources expand alongside the company’s success. Its mission remains to ensure that artificial general intelligence benefits all of humanity.

The more the business prospers, the greater the Foundation’s capacity to fund global initiatives.

An initial $25 billion commitment will focus on two core areas: advancing healthcare breakthroughs and strengthening AI resilience. Funds will go toward open-source health datasets, medical research, and technical defences to make AI systems safer and more reliable.

The initiative builds on OpenAI’s existing People-First AI Fund and reflects recommendations from its Nonprofit Commission.

The recapitalisation follows nearly a year of discussions with the Attorneys General of California and Delaware, resulting in stronger governance and accountability. With this structure, OpenAI aims to advance science, promote global cooperation, and share AI benefits broadly.

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

Rare but real, mental health risks at ChatGPT scale

OpenAI says a small share of ChatGPT users show possible signs of mental health emergencies each week, including mania, psychosis, or suicidal thoughts. The company estimates 0.07 percent and says safety prompts are triggered. Critics argue that small percentages scale at ChatGPT’s size.

A further 0.15 percent of weekly users discuss explicit indicators of potential suicidal planning or intent. Updates aim to respond more safely and empathetically, and to flag indirect self-harm signals. Sensitive chats can be routed to safer models in a new window.

More than 170 clinicians across 60 countries advise OpenAI on risk cues and responses. Guidance focuses on encouraging users to seek real-world support. Researchers warn vulnerable people may struggle to act on on-screen warnings.

External specialists see both value and limits. AI may widen access when services are stretched, yet automated advice can mislead. Risks include reinforcing delusions and misplaced trust in authoritative-sounding output.

Legal and public scrutiny is rising after high-profile cases linked to chatbot interactions. Families and campaigners want more transparent accountability and stronger guardrails. Regulators continue to debate transparency, escalation pathways, and duty of care.

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

OpenAI and Microsoft sign new $135 billion agreement to deepen AI partnership

Microsoft and OpenAI have signed a new agreement that marks the next phase of their long-standing partnership, deepening ties first formed in 2019.

The updated deal builds on years of collaboration in advancing responsible AI, positioning both organisations for long-term success while introducing new structural and operational changes.

Under the new arrangement, Microsoft supports OpenAI’s transition into a public benefit corporation (PBC) and recapitalisation. The technology giant now holds an investment valued at around $135 billion, representing about 27 percent of OpenAI Group PBC on an as-converted diluted basis.

Despite OpenAI’s recent funding rounds, Microsoft previously held a 32.5 percent stake in the for-profit entity.

The partnership maintains Microsoft’s exclusive rights to OpenAI’s frontier models and Azure API until artificial general intelligence (AGI) is achieved, but also introduces several new terms. Once AGI is declared, an independent panel will verify it.

Microsoft’s intellectual property rights are extended through 2032, including models developed after AGI with safety conditions. OpenAI may now co-develop certain products with third parties, while retaining the option to serve non-API products on any cloud provider.

OpenAI will purchase an additional $250 billion worth of Azure services, although Microsoft will no longer hold first-refusal rights for compute supply. The new framework allows both organisations to innovate independently, with Microsoft permitted to pursue AGI independently or with other partners.

The updated agreement reflects a more flexible collaboration that balances independence, growth, and shared innovation.

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

New ChatGPT model reduces unsafe replies by up to 80%

OpenAI has updated ChatGPT’s default model after working with more than 170 mental health clinicians to help the system better spot distress, de-escalate conversations and point users to real-world support.

The update routes sensitive exchanges to safer models, expands access to crisis hotlines and adds gentle prompts to take breaks, aiming to reduce harmful responses rather than simply offering more content.

Measured improvements are significant across three priority areas: severe mental health symptoms such as psychosis and mania, self-harm and suicide, and unhealthy emotional reliance on AI.

OpenAI reports that undesired responses fell between 65 and 80 percent in production traffic and that independent clinician reviews show significant gains compared with earlier models. At the same time, rare but high-risk scenarios remain a focus for further testing.

The company used a five-step process to shape the changes: define harms, measure them, validate approaches with experts, mitigate risks through post-training and product updates, and keep iterating.

Evaluations combine real-world traffic estimates with structured adversarial tests, so better ChatGPT safeguards are in place now, and further refinements are planned as understanding and measurement methods evolve.

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

Celebrity estates push back on Sora as app surges to No.1

OpenAI’s short-video app Sora topped one million downloads in under a week, then ran headlong into a likeness-rights firestorm. Celebrity families and studios demanded stricter controls. Estates for figures like Martin Luther King Jr. sought blocks on unauthorised cameos.

Users showcased hyperreal mashups that blurred satire and deception, from cartoon crossovers to dead celebrities in improbable scenes. All clips are AI-made, yet reposting across platforms spread confusion. Viewers faced a constant real-or-fake dilemma.

Rights holders pressed for consent, compensation, and veto power over characters and personas. OpenAI shifted toward opt-in for copyrighted properties and enabled estate requests to restrict cameos. Policy language on who qualifies as a public figure remains fuzzy.

Agencies and unions amplified pressure, warning of exploitation and reputational risks. Detection firms reported a surge in takedown requests for unauthorised impersonations. Watermarks exist, but removal tools undercut provenance and complicate enforcement.

Researchers warned about a growing fog of doubt as realistic fakes multiply. Every day, people are placed in deceptive scenarios, while bad actors exploit deniability. OpenAI promised stronger guardrails as Sora scales within tighter rules.

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

MLK estate pushback prompts new Sora 2 guardrails at OpenAI

OpenAI paused the ability to re-create Martin Luther King Jr. in Sora 2 after Bernice King objected to user videos. Company leaders issued a joint statement with the King estate. New guardrails will govern depictions of historical figures on the app.

OpenAI said families and authorised estates should control how likenesses appear. Representatives can request removal or opt-outs. Free speech was acknowledged, but respectful use and consent were emphasised.

Policy scope remains unsettled, including who counts as a public figure. Case-by-case requests may dominate early enforcement. Transparency commitments arrived without full definitions or timelines.

Industry pressure intensified as major talent agencies opted out of clients. CAA and UTA cited exploitation and legal exposure. Some creators welcomed the tool, showing a split among public figures.

User appetite for realistic cameos continues to test boundaries. Rights of publicity and postmortem controls vary by state. OpenAI promised stronger safeguards while Sora 2 evolves.

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

OpenAI rolls out pet-centric AI video features and social tools in Sora

OpenAI has announced significant enhancements to its text-to-video app Sora. The update introduces new features including pet and object ‘cameos’ in AI-generated videos, expanded video editing tools, social sharing elements and a forthcoming Android version of the app.

Using the new pet cameo feature, users will be able to upload photos of their pets or objects and then incorporate them into animated video scenes generated by Sora. The objective is to deepen personalisation and creative expression by letting users centre their own non-human characters.

Sora is also gaining editing capabilities that simplify the creation process. Users can remix existing videos, apply stylistic changes, and integrate social-type features like feeds where others’ creations can be viewed and shared. The Android app is noted as ‘coming soon’ which expands Sora’s accessibility beyond the iOS/web initial release.

The move reflects OpenAI’s strategy to transition Sora from an experimental novelty into a more fully featured social video product. By enabling user-owned content (pets/objects), expanding sharing functionality and broadening platform reach, Sora is positioned to compete in the generative video and social media landscape.

At the same time, the update raises questions around content use, copyright (especially when user-owned pets or objects are included), deepfake risks, and moderation. Given Sora’s prior scrutiny over synthetic media, the expansion into more personalised video may prompt further regulatory or ethical review.

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