# Deepfake İlgili Makaleler

HTX Haber Merkezi, kripto endüstrisindeki piyasa trendleri, proje güncellemeleri, teknoloji gelişmeleri ve düzenleyici politikaları kapsayan "Deepfake" hakkında en son makaleleri ve derinlemesine analizleri sunmaktadır.

3 Million-Follower Goddess Fully AI-Synthesized, Fake Orphanage, Cross-Border ‘Fake Charity’ Collapses Overnight

An Australian influencer named Lily Jay, with nearly 3 million Instagram followers, is at the centre of a major AI-fueled charity scam. Her 'Lily Jay Foundation' claimed to build orphanages, mosques, and deliver aid in places like Uganda and Gaza, using emotionally powerful videos of herself with children. However, an investigation by ABC NEWS Verify revealed these scenes—including a key video announcing an orphanage's opening—were entirely AI-generated, complete with telltale flaws like garbled text on clothing. The foundation's 'humanitarian award' was also faked using AI images. The foundation admitted in small print on its website that it is not a registered charity, avoiding financial transparency laws. While based in Kosovo, its operations and the influencer's lavish personal lifestyle raised further red flags. Following media inquiries, the foundation selectively hid donation buttons and removed fraudulent content for Australian visitors, but kept them active internationally. Experts warn this case is a textbook example of how AI can be weaponized to exploit public trust and generosity, creating believable but entirely fabricated narratives of aid. The incident highlights the growing danger of deepfakes in eroding trust within the charitable sector, urging donors to critically verify causes before giving.

marsbit07/21 12:57

3 Million-Follower Goddess Fully AI-Synthesized, Fake Orphanage, Cross-Border ‘Fake Charity’ Collapses Overnight

marsbit07/21 12:57

Shocking Mega-Scam: The 'Mysterious Lab' That Topped Global Charts Overnight Was Fake

On July 18th, a sensational announcement shook the AI community: a mysterious Chinese AI lab called "Basalt Labs" had seemingly released the "Monolith-1.0" model, topping major global benchmarks with claims of 1.6 trillion parameters and near-perfect scores. Its professional-looking website and technical paper fueled widespread excitement, raising questions about a sudden leap in China's AI capabilities. However, the story quickly unraveled. Developers discovered the model's Hugging Face repository contained duplicated weight files from a much smaller model. The impressive web demo was found to be a shell, secretly using DeepSeek's API for its responses. Prompt leaks further confirmed the deception. The mastermind, Max Scherf, later admitted it was an elaborate hoax, a "social experiment." He revealed creating the illusion by fine-tuning a small open-source model on leaked benchmark answers, fabricating all data, building a convincing website, and launching a viral marketing campaign. His goal was to expose the AI industry's vulnerabilities: an over-reliance on impressive benchmarks and hype, coupled with a lack of immediate, thorough scrutiny. Ironically, the scam highlighted the genuine strength of Chinese AI models like Qwen and DeepSeek, whose capabilities were credible enough to temporarily impersonate a "world-leading" model.

marsbit07/20 02:48

Shocking Mega-Scam: The 'Mysterious Lab' That Topped Global Charts Overnight Was Fake

marsbit07/20 02:48

DeepMind's Classic Masterpiece Crowned Again, ICML 2026 Awards Announced

ICML 2026 has announced its annual awards, with diffusion models and AI safety ethics taking center stage. The Outstanding Paper Award was shared by two diffusion model studies. One challenges a core assumption of diffusion language models (DLMs), arguing that their touted "arbitrary order generation" is a "flexibility trap" that harms performance. The other provides a high-accuracy sampling method, pushing the technical ceiling for diffusion models and log-concave distributions. A position paper winning the Outstanding Award raises a critical ethical concern: AI alignment research is unintentionally building a "censor's toolkit," where safety tools like RLHF can be repurposed for content control. Several papers received Honorable Mentions, spanning key areas: mapping where honesty emerges in RLHF-trained models, motion attribution in video generation, quantifying how much language models memorize, analyzing diffusion model consistency via random matrix theory, and providing a mathematical proof for the "grokking" phenomenon in a simple model. The Test of Time Award was given to DeepMind's 2016 seminal work "Asynchronous Methods for Deep Reinforcement Learning," recognizing the enduring impact of the A3C algorithm. Overall, the awards signal a shift in AI research from rapid expansion to deeper scrutiny—validating diffusion models as a major architectural contender while prompting serious ethical reflection within the safety community.

marsbit07/06 02:38

DeepMind's Classic Masterpiece Crowned Again, ICML 2026 Awards Announced

marsbit07/06 02:38

How to Detect AI-Generated Videos? A Review of Dynamic, Traceable, and Explainable Detection Systems

**How to Detect AI-Generated Videos: A Survey on Dynamic, Traceable, and Explainable Detection Systems** With rapid advances in AI video generation (e.g., Sora, Veo), creating highly realistic, multi-minute videos is now possible, widening the gap with detection research. Current AI video detection, often limited to unreliable binary classifications, is insufficient. This survey, accepted at ACL 2026, reframes the goal as **"factual fidelity verification"**—checking if a video's content (who, when, where, what) aligns with the real world perceptually and cognitively. It categorizes AI-generated videos into three paradigms: **Local Manipulation Videos (LMV**, e.g., face swaps), **Audio-Visual Editing (AVE**, e.g., lip-syncing), and **Generative Video Synthesis (GVS**, fully synthetic videos like Sora's). Detection challenges evolve from visual artifacts in LMV to multi-modal inconsistencies in AVE and higher-level world knowledge violations in GVS. The core proposal is a **Vision-Language Dual-View framework** with four hierarchical layers: 1. **Layer 1 (Intrinsic Visual Cues):** Analyzes low-level signal statistics, noise patterns, and physiological signals. 2. **Layer 2 (Spatiotemporal Consistency):** Checks for temporal coherence in object motion and scene dynamics. 3. **Layer 3 (Cross-Modal Consistency):** Verifies alignment between video, audio, and text within the video. 4. **Layer 4 (Language-Guided World-Level Reasoning):** Uses external knowledge, facts, and physical laws to judge semantic plausibility and factual correctness. The survey traces a shift in detection focus from lower layers (1 & 2) toward higher, language-involved layers (3 & 4). It also reviews evolving evaluation metrics and datasets tailored for each video paradigm. The conclusion advocates for a **dynamic, evidence-first detection system** that moves beyond simple classification. Future trustworthy detection requires combining visual evidence (from CV) with semantic reasoning and explanation (from NLP & multimodal AI), ultimately creating traceable and explainable judgments about a video's adherence to real-world constraints.

marsbit06/26 07:27

How to Detect AI-Generated Videos? A Review of Dynamic, Traceable, and Explainable Detection Systems

marsbit06/26 07:27

AI Has Started Forging Local Media

AI-Generated Fake Local Media Outlets Proliferate in the US A recent investigation by *The Florida Trident* exposed a network of fake local news websites, including the *South Florida Standard*, that are entirely generated and operated by artificial intelligence. These sites present fabricated editorial teams with AI-generated journalist photos, biographies, and bylines. Their content is not original reporting but consists of articles scraped from legitimate news outlets and rewritten by AI. The investigation traced the operation to Drew Chapin, a Philadelphia-based online reputation manager and convicted fraudster. Chapin admitted to running an "experiment" involving 17 such AI-driven news sites across the U.S., which generated over 3,500 URLs and attracted approximately 44,000 visitors. He demonstrated that a fully functional site, complete with a fake staff and content, could be created for just $10 and 15 minutes using AI prompts. These sites are part of a growing phenomenon known as "pink slime" journalism—low-quality, often deceptive outlets filling the vacuum left by the decline of local newspapers. Experts warn that such AI-forged media not only deceive readers but also threaten legitimate news organizations by scraping their content. A greater danger lies in future AI models potentially being trained on this fraudulent, low-quality information, polluting the digital information ecosystem.

marsbit05/16 03:40

AI Has Started Forging Local Media

marsbit05/16 03:40

This Time, OpenAI Eliminated 90% of Human Designers

OpenAI's latest release, GPT-Image 2, marks a paradigm shift in AI-generated imagery, moving beyond aesthetic quality to logical reasoning and contextual understanding. The model introduces a "thinking mode," where it performs background reasoning—such as mathematical calculations and geographic knowledge—before generating images. This enables highly accurate and context-aware outputs, like a livestream overlay showing precise distance metrics or a brand-aligned poster design. The model excels in rendering Chinese text with remarkable accuracy and aesthetic quality, a significant improvement over previous versions. It supports multi-turn conversational editing via the new Responses API, allowing iterative refinements similar to chatting with a large language model. While GPT-Image 2 demonstrates unprecedented capabilities in commercial applications like marketing material and illustration—potentially displacing many human designers due to its cost efficiency—it still has limitations. Minor artifacts in fine text details persist, and complex prompts can cause extended processing times. Additionally, the technology raises ethical concerns around deepfakes and digital trust. Overall, GPT-Image 2 transitions AI image generation from a novelty to a powerful production-ready tool, redefining industry standards and pushing the boundary of what’s possible in visual AI.

marsbit04/23 10:03

This Time, OpenAI Eliminated 90% of Human Designers

marsbit04/23 10:03

Hong Kong Crypto Scam Shock: Woman Loses Nearly $1 Million As AI Fraud Surges

Hong Kong Crypto Scam Shock: Woman Loses Nearly $1 Million As AI Fraud Surges A Hong Kong woman lost HK$7.7 million (approximately $982,000) after being lured into a fraudulent cryptocurrency investment platform. The scam began on Telegram, where a person posing as an investment expert promoted an AI-powered trading strategy with guaranteed returns. After making 17 transfers of USDT and Ethereum, the victim was unable to withdraw her funds. Hong Kong Police have reported a significant surge in such online investment fraud, with over 80 cases recorded in a single week, totaling losses of nearly HK$80 million. Authorities warn that scammers are increasingly using sophisticated tactics, including promises of "AI trading" and "guaranteed profits," to appear credible. This follows a similar incident last month where a retiree lost HK$6.6 million in a multi-stage scam. Police urge the public to be cautious of unsolicited investment advice and to verify platforms through the official CyberDefender service before transferring any money. They emphasize that no legitimate investment can guarantee returns. Investigations are ongoing.

bitcoinist04/19 14:01

Hong Kong Crypto Scam Shock: Woman Loses Nearly $1 Million As AI Fraud Surges

bitcoinist04/19 14:01

$20 for a Face: The Underground Business of Crypto KYC

Crypto KYC Bypass: A $20 Underground Industry Despite stringent KYC (Know Your Customer) requirements from major crypto exchanges, a thriving underground market exists to bypass these checks for as low as $20. Users often face geo-blocks or lengthy verifications, preventing access to services. This has fueled demand for illicit KYC services. Reports indicate over 500,000 participants in underground KYC markets, with more than 1 million listings selling verified profiles from platforms like Coinbase and Kraken. These accounts often include real personal data, sometimes without the original owners' knowledge. Fraud techniques have evolved, including deepfake attacks (up 2000% in three years), screen-based spoofing, and AI-generated fake documents. The virtual currency sector is the primary target, accounting for over 78% of KYC attacks. An investigation into a Telegram-based KYC vendor revealed a TRON address with over $59,000 in USDT from 600 transactions over two years, all eventually transferred to an OKX hot wallet. An interview with a KYC service provider, "Maoli," who operates in Chinese-speaking regions, detailed the process: clients pay for accounts verified by "foreigners" recruited globally, often from lower-income regions, who perform the KYC steps for a small fee. These accounts are sold with warnings against holding large funds due to fraud risks and potential reclaiming by the original identity owners. Maoli described the business as a "three-way win": users gain access, exchanges get user numbers, and he profits. However, this ignores the victims of identity theft whose data is used without consent. The KYC system, while intended for security, functions as a permeable barrier, with a vast shadow economy ensuring access for those willing to pay.

marsbit03/30 07:36

$20 for a Face: The Underground Business of Crypto KYC

marsbit03/30 07:36

NVIDIA Starts Installing Chips on Roads | Rewire Evening News Update

NVIDIA CEO Jensen Huang announced at GTC that the company's data center orders for Blackwell and Vera Rubin platforms are projected to exceed $1 trillion by 2027, doubling last year's estimates. He emphasized that computing demand will far surpass this figure. Beyond data centers, NVIDIA is expanding its autonomous driving ecosystem, adding BYD, Geely, Nissan, and Isuzu to its Drive Hyperion platform. A partnership with Uber aims to deploy robotaxis in Los Angeles and San Francisco by early 2027, expanding to 28 markets by 2028—a moment Huang calls "the ChatGPT moment for autonomous driving." In related news, Uber co-founder Travis Kalanick revealed his stealth robotics startup, Atoms, after eight years of operation. The company focuses on automating physical infrastructure, mining, and robotic platforms. Kalanick is reportedly acquiring autonomous driving firm Pronto, with Uber's support, signaling a strategic re-entry into automation. Meanwhile, Murata Manufacturing, the world's largest MLCC supplier with over 40% market share, raised prices for AI server and automotive-grade components by 15-35%, effective April 1. This marks its first major price hike in three years and highlights hidden cost pressures in AI infrastructure supply chains. The SEC is also considering allowing public companies to switch from quarterly to semi-annual financial reporting, reducing compliance costs and potentially benefiting tech firms making long-term AI investments. Additional updates include Alibaba providing employees with free AI tool tokens, FDIC moving to exclude stablecoins from deposit insurance, deepfake misinformation spreading during the Israel-Hamas war, and Picsart launching an AI Agent marketplace for creators.

marsbit03/17 19:08

NVIDIA Starts Installing Chips on Roads | Rewire Evening News Update

marsbit03/17 19:08

VeryAI Raises $10M to Launch Palm-Scan Identity Verification on Solana

VeryAI has raised $10 million in a seed funding round led by Polychain Capital, with participation from Berggruen Institute and Anagram. The company is launching a hardware-free palm-scan identity verification system to combat AI-generated fraud and deepfakes. The system uses smartphone cameras to capture palm biometrics, claiming a false acceptance rate of 1 in 10 million for one hand and 1 in 100 trillion for both hands. Built on Solana, the platform uses Zero Knowledge Proofs and the Solana Attestation Service to ensure privacy and interoperability. VeryAI aims to provide businesses like crypto exchanges and fintech companies with a secure, private, and accessible identity verification solution.

TheNewsCrypto03/13 00:22

VeryAI Raises $10M to Launch Palm-Scan Identity Verification on Solana

TheNewsCrypto03/13 00:22

活动图片

Trading Strategiesright-arrow

Ethereumright-arrow