chatgpt Timeline - FediScanner

En su artículo, Jorge Majfud explora el papel de ChatGPT en el proceso creativo y la escritura. Destaca que, aunque estas herramientas pueden ayudar a generar ideas, la humanidad y la autenticidad siguen siendo esenciales. "La máquina puede generar discursos, pero el sentido y el contexto son humanos."

¿Cómo se hoy en día la "esencia única de nuestras experiencias" ?

https://www.nodal.am/2025/11/chatgpt-por-que-para-que-y-para-quien-escribimos-por-jorge-majfud/

Show Original Post

Report

21.01.2026 20:00
InfobloxThreatIntel (@InfobloxThreatIntel@infosec.exchange)

Like many research teams, we've been tracking the various flavors of 'ClickFix' attacks, but this example really stands out because it abuses a seemingly legitimate ChatGPT output:
https://urlscan.io/result/019b088f-8744-760a-8da0-abd870fdf64e/
ChatGPT allows users to share their conversations publicly on chatgpt.com. In this case, the shared conversation appears to be helpful instructions on "How to clear disk space on macOS," except the commands presented are malicious and would trick users into downloading and executing malware from putuartana[.]com.
The malicious command uses curl with flags (-fsSL) designed to:
• Download silently without user-visible errors
• Pipe the payload directly to bash for immediate execution
• Leave no trace on disk, running entirely in memory
To generate this output on ChatGPT's platform, the threat actor likely employed prompt injection techniques to manipulate the LLM into producing the malicious tutorial. Once created, this conversation could be shared via social media, forums, or SEO-optimized pages, directing victims to legitimate chatgpt.com infrastructure hosting the attack.
Both the ChatGPT shared page and the malware (hosted on putuartana[.]com) have since been taken down. However, we expect this won't be the last time threat actors weaponize LLM outputs in this manner.
This technique is particularly dangerous because:
• Content is hosted on a trusted domain (chatgpt.com)
• It appears professionally formatted and helpful
• Users are conditioned to trust AI-generated instructions
• No traditional phishing indicators (suspicious URLs, poor grammar) are present
The lesson here? Educate users to never to run unfamiliar terminal/PowerShell commands from untrusted sources, even when hosted on seemingly legitimate platforms.
IOC: putuartana[.]com
#dns #threatintel #threatintelligence #cybercrime #cybersecurity #infosec #infoblox #infobloxthreatintel #chatgpt #ai #llm #clickfix

Show Original Post

Report

21.01.2026 19:07
BrentD (@BrentD@techhub.social)

The game is changing for Advertisers again. OpenAI is preparing to release Ads in ChatGPT "conversations", possibly in Feb.

Advertisers will initially pay per impression (PPM), and not per click (PPC) as is common with web-based advertising. Ads will be placed at the end of answers in ChatGPT, clearly labelled as “Sponsored” and separated from what it calls the “organic” answer. Users will have the option to indicate if they want to see more or less Ads in the future.

Initially OpenAI will not provide self-service tools for buying and managing ads during the test phase.

Marketers will need to figure out how to leverage conversational interfaces to support long-term brand building alongside performance outcomes. The challenge for OpenAI is to find an acceptable balance of user trust, transparency and creativity, and whether advertising is positioned as a value add to the experience rather than a distraction from it. https://digiday.com/marketing/as-openai-gears-up-to-launch-chatgpt-ads-marketers-try-to-keep-up/ #OpenAI #ChatGPT #Ads #Marketing #DigitalMarketing #AI

Show Original Post

Report

1 ...616 617 618 619 620 621 622 623 624 625 626 ...1664