Business Ethics & Governance News: August 2026
Regulators spent the last fortnight of July on two themes that sit at the centre of corporate ethics work: enforcing the rules on the largest platforms, and making the use of artificial intelligence visible. The European Commission handed Google its first fine under the Digital Markets Act, the AI Act's transparency duties for AI-generated content reached their application date, and lawmakers in Washington proposed a single federal rulebook for the most powerful AI models. Here is what changed and why it matters for compliance and governance teams.
Commission fines Google 890 million euros under the Digital Markets Act
On 23 July 2026 the European Commission fined Google 890 million euros for breaching the Digital Markets Act, the first penalty the Commission has imposed on the company under that law. The Commission found two failings: Google gave preferential treatment to its own services, such as shopping, hotels, transport and sports results, over rival services in Google Search, and it stopped app developers on Google Play from freely steering users to cheaper offers outside the store. The total splits into 460 million euros for the search self-preferencing and 430 million euros for the app-store steering restrictions.
Google must bring the conduct to an end within 60 days or face periodic penalty payments. For governance and competition-compliance teams, the decision is a reminder that the DMA carries real financial consequences for designated gatekeepers, and that "self-preferencing" and how a platform routes users to third parties are now enforcement priorities rather than abstract principles.
Source: European Commission
AI Act transparency duties for AI-generated content start to apply
From 2 August 2026 the transparency obligations in Article 50 of the EU AI Act apply. They cover four situations: providers must tell people when they are interacting with an AI system unless it is obvious; providers must mark AI-generated audio, image, video and text so it is machine-readable and detectable as artificially produced; deployers must inform people who are subject to emotion-recognition or biometric-categorisation systems; and deployers must disclose deepfakes and AI-generated text published on matters of public interest.
There is a short reprieve on one point. For systems already on the market before 2 August 2026, the duty to mark synthetic content is delayed until 2 December 2026. The practical task for firms that use AI to write marketing copy, generate images or run chatbots is to confirm that such outputs are labelled and that an internal owner is named for the disclosure. A voluntary Code of Practice can guide the approach, but it does not remove the underlying legal duty.
Source: European Commission
US lawmakers introduce the FRONTIER Act for federal AI oversight
On 23 July 2026 Representatives Jay Obernolte and Lori Trahan introduced the bipartisan FRONTIER Act, which would set a national framework for the development and deployment of the most capable AI models. The bill applies only to "frontier" models, anchored to a compute threshold of more than 10 to the power of 26 training operations, so that everyday systems fall outside it. Large developers would face tiered duties including transparency reports, risk-management plans, independent third-party audits and reporting of critical safety incidents.
The bill would also give federal authorities an emergency power to restrict a frontier model on grounds of imminent catastrophic risk, backed by civil penalties of up to 10 million dollars per violation per day. Its sponsors present it as a way to replace a patchwork of state rules with one federal standard. It is only a proposal for now, but it signals the direction of US AI governance and is worth tracking for firms that build on or buy frontier models.
Source: US House of Representatives