AI Governance Is No Longer Optional: What Privacy Employers Are Really Asking For
If you’ve been reading our quarterly Privacy Job Repor... ...
The team at Privacy 108 has collated news from around the world covering artificial intelligence laws and regulations, legal cases, security, and general AI news. You can find our summaries and links to the original coverage right here.
Anthropic CEO Dario Amodei penned an essay called “Policy on the AI Exponential”, calling on AI regulation to move faster to match the ‘lightning pace’ of the industry, coming alongside policy proposals on AI testing and employment disruption. Amodei said Claude Mythos Preview’s hacking risks are a major turning point, making frontier models “tools of global and national strategic consequence.” The CEO wants regulators to have the power to ground frontier models, with Amodei proposing systems independently screened across four risk areas. Other asks include faster approval of AI-designed drugs, limits and bans on autonomous weapons, and stronger advanced chip export controls. Amodei’s thought piece is here.
The Wall Street Journal reports on how Anthropic’s Claude Mythos model is pushing US President Donald Trump’s administration to increase AI safety considerations instead of focusing on innovation and global leadership. Vice President JD Vance recently urged major AI developers to work together to combat safety and security concerns Mythos’ cybersecurity vulnerability detection capabilities have raised. The White House is also considering executive actions to establish more government oversight of advanced AI systems before they are sent to market. More here.
Comprehensive AI regulation, the Great American Artificial Intelligence Act of 2026, has been introduced with bipartisan support, Politico reports. The bill introduces obligations for companies, requiring them to address significant safety risks and complete third-party audits to ensure AI systems are secure. Americans for Responsible Innovation President Brad Carson raised concerns about the bill’s ability to pre-empt states’ AI laws, noting it would be a “generational mistake” to override states’ current AI efforts. More here.
An investigation by Canada’s Office of the Privacy Commissioner and provincial privacy authorities from Alberta, British Columbia and Quebec found OpenAI’s AI data training practices allegedly violated the relevant federal and provincial privacy statutes, including the Personal Information Protection and Electronic Documents Act. The probe found alleged shortcomings related to OpenAI’s data collection and consent practices, as well as issues around data subject access requests. More here.
The agreed amendments to the AI Act include an explicit ban on so-called nudifier apps that depict an identifiable person’s intimate parts or show them engaged in sexually explicit activities, or that generate child sexual abuse material. That ban will come into force on 2 December. The European Commission is already investigating Elon Musk’s X and Grok AI over such content under a separate regulation. More here.
Governor Kim Reynolds, R-Iowa, signed Iowa Senate File 2417, a law aiming to bolster protections for minors under age 18 when interacting with AI chatbots. The law, which takes effect 1 July 2027, requires AI chatbot providers to remind underage users that a bot is not human and implement safeguards to prevent tools from engaging in conversations that could be considered inappropriate or harmful. The law provides for attorney general rulemaking and civil penalties of up to USD1,000 per violation. More here and legislation is here.
Although recreating an Australian politician’s likeness is not “automatically illegal”, takedown orders, penalties under Australian Consumer Law, or defamation claims may follow if a social media post misleads viewers, lawyers have warned. More here.
The eSafety Commissioner has been busy with the announcement of formal enforcement action against an unnamed AI “nudify” platform for alleged nonconsensual exploitation and inadequate safeguards to prevent underage users from accessing explicit images. Inman Grant said protecting children from “exposure to harmful and age-inappropriate content is a core priority, and we expect providers offering services to and profiting from Australian end-users, including children — no matter where they are based — to follow Australian law.” More here.
A California jury dismissed Elon Musk’s lawsuit against his former business partner, OpenAI CEO Sam Altman. The jury said Musk waited too long to file his lawsuit. Musk alleged that Altman and co-founder Greg Brockman breached their charitable trust by abandoning OpenAI’s founding mission and profiting from that decision. The unanimous decision of the nine-member advisory jury took less than two hours. They said Musk exceeded the statute of limitations when he launched his case in 2024.
Florida Attorney General James Uthmeier filed a lawsuit against OpenAI in state court alleging the company’s ChatGPT poses a danger to children and that it does not warn the public about the chatbot’s risks, The New York Times reports. The complaint alleges OpenAI built “a dangerous online product where harmful information such as tips on eating disorders, self-harm and mass murder are readily available, including to young children. More here.
President Donald Trump signed his anticipated executive order addressing cybersecurity concerns associated with frontier AI. The order loosens some of the expected requirements imposed on AI developers compared to the previously leaked executive order that was never issued, such as shortening the voluntary government review window for new models from up to 90 days to 30 days. Additionally, the order establishes an entity within the Department of the Treasury that would help coordinate patching efforts for software vulnerabilities. More here.
The US National Institute of Standards and Technology’s Center for AI Standards and Innovation and the UK AI Security Institute each announced new partnerships with major AI developers that will allow pre-deployment evaluations and targeted research of frontier models to ensure model security and safety. CAISI entered into agreements with Google DeepMind, Microsoft and xAI, while the UK AISI secured a deal with Microsoft. More here.
In an interview with Reuters, European Securities and Markets Authority Chair Verena Ross said the risk and speed of cyberattacks powered by AI are growing. Ross said her office has been contacting financial institutions to assess the resiliency of their cyber defences amid advances in AI, and called for improved collective capability at both national and EU level to oversee critical third-party providers. More here.
Against the backdrop of uncertainty in the EU over gaining access to Anthropic’s Claude Mythos AI model, OpenAI has offered the European Commission access to its latest AI model that purports to similarly discover cybersecurity vulnerabilities. During a May hearing before the European Parliament’s Committee on the Internal Market and Consumer Protection, officials from the Commission, the EU AI Office and the European Union Agency for Cybersecurity discussed the cyber risks posed by frontier AI models, as well as opportunities to harness them for defensive purposes. More here.
Anthropic released AI model Claude Fable 5, a “Mythos-class” model, to enterprise and paid customers two months after it initially released Mythos to a limited number of users due to its potential to exploit software vulnerabilities, CNBC reports. The company said Claude Fable 5 demonstrates “exceptional performance” for software engineering and knowledge work tasks and scores upward of 10% higher than Claude Opus 4.8 on some benchmarks. This is why Anthropic had to implement additional guardrails to prevent misuse. If a user asks a high-risk question, like how to make ricin, a toxin, for instance, the model will block its response and fall back to Claude Opus 4.8 to deliver a safe answer. More here.
Anthropic called for pausing global AI development out of concern frontier AI models are advancing so rapidly they could potentially improve themselves without direct human intervention, The Wall Street Journal reports. The company said frontier models appear to be undergoing a process of “recursive self-improvement.” The company proposed establishing a global structure on slowing development and collaborating with partners to ensure malicious and irresponsible actors do not weaponise the most cutting-edge models capable of recursive self-improvement amid a potential wider pause on development. More here.
Experts at the International Monetary Fund (IMF) see the rapid development of AI-powered cyberattack tools as a growing threat to the global financial system. Central to their analysis is, among other things, the AI model Claude Mythos Preview from Anthropic, whose controlled release the analysts regard as a warning signal. More here.
Pope Leo has issued a papal declaration addressing the ethical dimensions of artificial intelligence. The declaration follows a series of Vatican statements on AI governance and calls for technology to serve human dignity and the common good, adding a significant moral voice to global AI governance debates. More here.
The Australian Signals Directorate and Australian Cyber Security Centre issued joint guidance for organisations adopting agentic AI solutions. The guidance is geared toward the public sector, critical infrastructure IT systems and industry stakeholders to help them understand the key security challenges and risks posed by agentic AI. The ASD and ACSC recommend adopting agentic AI “incrementally,” while “limiting them to low-risk tasks.” More here.
OpenAI published a proposed blueprint for the US to build resilient institutions to safely govern frontier AI. The blueprint contains a three-part strategy: building a national framework reflecting the emerging consensus on state frontier AI safety law, strengthening the National Institute of Standards and Technology’s Center for AI Standards and Innovation as the US government’s primary institution for frontier AI safety, and developing a broader resilience plan across government to address the national security and public safety challenges posed by frontier AI. More here.
The China-based company behind the “Great Firewall” software, Geedge Networks, is reportedly developing technology that uses AI to analyze location data and internet use to assess the likelihood of individuals becoming a political dissident, according to leaked documents obtained by The New York Times. Researchers at Vanderbilt University found Geedge, working with government-backed research entity MESA Lab, was developing AI technology that built profiles of Chinese citizens to evaluate if they may harbor anti-government sentiment. More here.
OpenAI is urging the G7 Leaders’ Summit to collaborate on an international standard of children’s AI protections and establish an “international youth safety institute.” The company said the G7 Summit could provide an opportunity for countries to discuss AI age-related protections such as parental controls, content restrictions and transparency. OpenAI said AI companies “should publish safety policies that explain what safeguards are in place to protect young people, parental tools available, and how protections are updated as risks evolve. This transparency builds trust and helps parents make informed choices.” More here.
The UK Information Commissioner’s Office issued a statement following the government’s announcement about the launch of its advisory AI Growth Lab. The lab will offer AI developers and adopters practical information on how existing laws apply to AI. “It’s essential that legal firms and businesses can navigate data protection requirements with confidence as they seek to make legal services more efficient and effective for people,” ICO Executive Director for Regulatory Risk and Innovation William Malcolm said. More here.
Canada announced its AI for All strategy that aims to bolster the economy by accelerating AI training for crucial industries and guiding organizations’ AI deployment. IAPP Staff Writer Lexie White unpacks the initiative’s efforts to strengthen AI innovation and advance Canada’s place in the AI race. More here.
The European Commission published its final voluntary Code of Practice for marking and labelling AI-generated content. The code establishes practical steps for generative AI developers and deployers to begin complying with the AI Act’s transparency obligations beginning 2 Aug. From that date, the AI Act will require labelling of deepfake content and AI-generated text, and a clear notice must be given to users engaging with an AI chatbot. More here.
India’s Ministry of Electronics and Information Technology proposed adding stricter rules surrounding the labelling of AI-generated content, The Economic Times reports. The proposed regulation would amend the Information Technology (Intermediary Guidelines and Digital Media Ethics Code) Rules and require all AI-generated content to be labelled as such throughout the entire duration of its display. More here.
"*" indicates required fields
"*" indicates required fields
Privacy 108 collects your name and email to send you our newsletter. If you do not provide this information, we will be unable to send it to you. We may use third-party service providers (such as email marketing platforms) to distribute our communications. Some providers may store information overseas, including in the United States. For more information about how we handle your personal information, including how to access or correct it or make a complaint, please see our Privacy Policy or contact us at hello@privacy108.com.au. You can unsubscribe at any time using the link in our emails or by contacting hello@privacy108.com.au.