AI-generated deepfakes pose a challenge not just to individuals but to institutional trust systems like open-source intelligence (OSINT). Examine with a recent example.

A deepfake is synthetic media that appears reasonably authentic — what India's amended IT Rules, 2021 call "synthetically generated information" [2]. Its deeper danger lies less in personal defamation than in corroding the evidentiary systems on which courts, humanitarian agencies and investigators rely.

The recent trigger: Google Earth's AI imagery (2026)

• On 30 July 2026, Google added an AI "create image" tool to Google Earth, letting users generate visuals layered over real satellite views [1].
• Within roughly a day, open-source investigators produced convincing fakes — a nuclear plant in Iran, refugee camps on the US–Mexico border, plane crashes and blast craters — forcing Google to roll back the feature pending "stronger guardrails" [1].

Harm at the individual level

• Impersonation and fraud: cloned voices and faces enable financial scams, non-consensual imagery and reputational damage.
• Remedies remain individual and reactive — grievance redressal and shortened takedown timelines under the amended IT Rules [2].

Harm to institutional trust systems

• OSINT credibility: satellite imagery functions as near-objective evidence; UNOSAT analysis substitutes for field verification in conflict zones inaccessible to humanitarian teams and supports international justice proceedings [4].
• Grounded forgery: because fakes are built on genuine terrain and buildings, they are internally consistent and hard to refute [1].
• Liar's dividend: authentic evidence of atrocities or disasters can be dismissed as AI-made, paralysing accountability.
• Low entry barrier: misuse no longer needs state-grade capability — casual users sufficed [1].

The governance gap

• A "ship-then-patch" rollout showed pre-launch red-teaming lagging behind capability [1].
• Fixes exist: machine-readable marking and provenance for AI outputs under the EU AI Act's transparency obligations [3], and mandatory labelling with metadata embedding under India's 2026 IT Rules amendments [2].

The episode shows that deepfake regulation must protect not only persons but public evidentiary infrastructure. Provenance-by-design, mandatory pre-deployment audits for tools built on trusted geospatial data, and capacity-building for verification bodies can preserve technology's promise while safeguarding institutional trust.