AI Voice-Cloning Scams Cost Americans Billions as Fraud Surges

·3 views

Scammers are using AI voice-cloning technology to impersonate loved ones or colleagues in fake crisis calls, tricking victims into sending money urgently and secretly. A short audio sample scraped from social media or voicemails is enough for widely available, low-cost tools to generate a convincing voice imitation. The US FTC recorded approximately $3.5 billion in impostor scam losses in 2025, while the FBI tracked around $893 million in AI-related fraud in its first-ever dedicated category. Deloitte projects generative-AI-enabled fraud in the US could reach $40 billion by 2027, up from roughly $12 billion in 2023. Security experts stress that the most effective defence remains behavioural — verifying any urgent, secretive request for money through a separate, trusted communication channel.

Read the full story at DEV Community

This is an AI-generated summary. ShortSingh links to the original source for the complete article.

Discussion (0)

Dev.to Writer Opens Up About Losing Passion for Tech and Coding

A software developer has shared a candid reflection on the Dev.to platform, expressing doubts about whether tech is the right field for them. Writing for the site's 2025 New Year's Writing Challenge, the author described struggling with the same professional concerns months later. They noted that the issue is not the difficulty of the work but a lack of measurable progress and personal goals within the industry. The developer said they have gradually lost their passions since graduating college and can no longer recall enjoying coding. Rather than withdrawing, they chose to share their experience publicly and invited others in the tech community to discuss how they are feeling about the field.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Google: AEO and GEO Are Part of SEO, Not Separate AI Search Strategies

Google has clarified that answer engine optimization and generative engine optimization are not disciplines separate from standard SEO, but extensions of it. The company states that the same foundational SEO practices — quality content, clear site structure, and crawlability — apply to its generative AI features like AI Overviews and AI Mode. Google's official guidance discourages businesses from building isolated optimization programs for AI search, urging them instead to integrate AI visibility into existing SEO workflows. Publishers are advised to use Search Console's Generative AI performance reporting to track visibility across AI-driven search surfaces. Google emphasizes that substantive, original content and technical accessibility remain the core requirements, rather than superficial tactics like question-and-answer reformatting or templated pages.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

How to Validate Gateway Tokens Using JWKS Caching and Key Rotation in Node.js 20

A technical guide published on DEV Community outlines best practices for validating e-commerce gateway tokens using public JWKS keys in Node.js 20, keeping private keys solely with the issuer. The approach uses an in-memory JWKS cache with a stale-while-refresh strategy, serving cached keys during their TTL and triggering a single refresh only when a key ID is missing. A hard ceiling on stale data prevents outdated keys from persisting through rotation events, while a retry-aware fetcher handles rate-limited identity provider responses gracefully. The guide emphasizes separating CAPTCHA verification from JWT authentication, treating each as a distinct, independently logged state transition to avoid ambiguous error signals. On retrieval failures or persistent rate limits, the recommended policy is to reject the request with a generic auth failure, emit an internal alert, and expose diagnostic metadata only to operators — never to public responses.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Engineer shares 8 hard blockers for reviewing AI-generated pull requests

A software engineer writing for DEV Community has outlined a strict eight-blocker rubric for evaluating AI-generated pull requests, arguing that fluent code and passing linters mask deeper risks. The framework focuses on issues that can cause outages, data corruption, access leaks, or irreversible changes — not style or naming preferences. Each blocker requires a cited file path and line number, and any PR carrying an unresolved blocker cannot be merged without written acknowledgment of the risk. Key examples include missing feature flags on irreversible migrations and silent breaking changes to API contracts, such as quietly converting an integer field to a formatted string. The author emphasizes that a clean run should produce silence, not suggestions, and that review tools which flag non-issues get ignored within days.

0 comments Read more at DEV Community