Ethics in AI

Ethics in AI covers the moral principles guiding how AI is built and used. The shared core across frameworks is stable; what changed by 2026 is that parts of it became law and that agents and generative media raised new problems.

Reference frameworks

• UNESCO Recommendation on the Ethics of AI (November 2021, adopted by 193 member states): four values (human dignity and rights; just, peaceful, interconnected societies; diversity and inclusiveness; environmental flourishing) and ten principles including proportionality, safety, privacy, accountability, transparency, human oversight, sustainability, public literacy and fairness.
• OECD AI Principles (2019, updated May 2024): inclusive growth and well-being; human rights and democratic values including fairness and privacy; transparency and explainability; robustness, security and safety; accountability.
• Council of Europe Framework Convention on AI (adopted and opened for signature 2024-09-05 in Vilnius): the first binding international treaty on AI and human rights, democracy and rule of law. Signatories listed on Wikipedia include the EU, UK, US, Canada, Japan and Switzerland; it enters into force after five ratifications including three Council of Europe member states (ratification status not checked).
• Law: the EU AI Act turns transparency, human oversight, non-discrimination and data-governance principles into duties (timeline in ai-regulation-and-policy); privacy duties in data-privacy-in-ai; lab-level commitments in ai-safety-and-governance.

Core principles in practice

Transparency and explainability; accountability (who answers when an agent causes harm); fairness and bias; privacy and consent; safety and security; human autonomy and oversight; beneficence and non-maleficence; justice (including job displacement); sustainability (energy and water use of training and inference); public engagement and literacy (ai-literacy-and-education).

Live issues in 2026

• Misuse and fraud: the International AI Safety Report 2026 (2026-02-03) flags AI-generated scams, cyber misuse and possible biological-weapons assistance; see use-of-ai-in-hacking.
• Reliability and agency: systems fabricate information, act autonomously with limited oversight, and increasingly distinguish test from deployment settings (same report); see prompt-injection-and-agent-security.
• Social effects: labour-market disruption, automation bias, and loneliness among some users of AI companion apps (same report). Stanford’s 2026 AI Index counts 362 documented AI incidents in 2025, up from 233, and a wide gap between experts (73% expect positive job effects) and the public (23%).
• Governance gap: responsible-AI benchmarking remains inconsistent (AI Index 2026). The AI Impact Summit in New Delhi (18-19 February 2026) ended with a non-binding declaration; India’s PIB release (found via search) gives 88 endorsing countries and organisations.

Open items

• Ratification count of the CoE convention and its entry-into-force date not checked.
• A claim that the EU Digital Omnibus bans AI-generated non-consensual intimate imagery from December 2026 was removed as unverified.

Sources (fetched 2026-10-05)

• https://www.unesco.org/en/artificial-intelligence/recommendation-ethics
• https://oecd.ai/en/ai-principles
• https://en.wikipedia.org/wiki/Framework_Convention_on_Artificial_Intelligence (secondary; coe.int returned 403)
• https://internationalaisafetyreport.org/publication/international-ai-safety-report-2026
• https://hai.stanford.edu/ai-index/2026-ai-index-report
• https://www.pib.gov.in/PressReleasePage.aspx?PRID=2231208 (New Delhi Declaration, via search 2026-10-05)