Lorenzo Arvanitis

student @harvardmed interested in cardiology.
formerly @NewsGuardRating studying misinformation.

[McKenzie Sadeghi @SadeghiMckenzie]

NewsGuard uncovered a network of nearly 1,500 fake accounts on Facebook and X that appears to be engaged in a coordinated effort to promote pro-India, AI-generarted content, via @Ddimitriadis_ https://t.co/3FMXMGOJfV

[Lorenzo Arvanitis @ZoArvanitis]

The results show that the chatbot could be easily used to spread misinformation at scale, in violation of OpenAI’s Usage Policies prohibiting the use of its services for “disinformation.”

Read the report on @NewsGuardRating's site.

[Lorenzo Arvanitis @ZoArvanitis]

OpenAI recently bragged that the new version of its AI model GPT-4 scores in the 90th percentile on the bar exam. But @SadeghiMckenzie, @jbruce218, and I found that ChatGPT-4 scored 0% in an exercise assessing its ability to avoid spreading misinformation.
https://t.co/Wutt76mmMR

[Lorenzo Arvanitis @ZoArvanitis]

Last week, we ran the same exercise on ChatGPT-4, using the same 100 false narratives and prompts. ChatGPT-4 responded with false and misleading claims for all 100 of the false narratives — and was more thorough and persuasive in advancing those false narratives than ChatGPT-3.5.

[NewsGuard @NewsGuardRating]

The number of sites identified by NewsGuard as spreading Russian disinformation and profiting from programmatic advertising has nearly tripled in the one year since Russia invaded Ukraine

https://t.co/3biVcZxtnQ

[NewsGuard @NewsGuardRating]

NEW: NewsGuard’s @ZoArvanitis gives an excellent overview with @FerretScot about our recent reporting on how ChatGPT could be used to spread misinfo.

🎧 Listen 2:30-14:20 🎧

[Lorenzo Arvanitis @ZoArvanitis]

@gloquenzi certo! [email protected]

[Jack Brewster @jbruce218]

New: @NewsGuardRating attempted to get ChatGPT to advance false information using its database of myths

The results confirm fears: For 80% of NewsGuard’s prompts, ChatGPT provided narratives containing false info

@ZoArvanitis, @SadeghiMckenzie & me

https://t.co/hEQJJ6Dnrg

[NewsGuard @NewsGuardRating]

.@GridNews cites NewsGuard’s reporting about how political orgs - posing as newsrooms - pushed their agendas on Facebook, and what it could mean with political advertising returning to Twitter.

[NewsGuard @NewsGuardRating]

NewsGuard found that TikTok is hosting hundreds of videos using violence and music to celebrate the Wagner Group, a notorious Russian mercenary outfit.

Read our exclusive report:
https://t.co/o7UccRJn8N

[Jack Brewster @jbruce218]

New: It's early yet, but misinformation superspreaders are enjoying an engagement boost in the post-Musk era, a @NewsGuardRating investigation finds.

https://t.co/Y0dDjzhU2F

[Lorenzo Arvanitis @ZoArvanitis]

New from @SadeghiMckenzie and me, in @Newsweek, tracing the evolution of the Paul Pelosi conspiracy and what it tells us about the state of our information ecosystem:

https://t.co/w0tKKHo87P

[NewsGuard @NewsGuardRating]

NEW: @Local12 spoke with NewsGuard’s @ZoArvanitis about “pink slime” newspapers showing up in SW Ohio.

[NewsGuard @NewsGuardRating]

Posing as local news outlets, partisan groups have exploited Meta’s hyper-targeting tools and porous ad policies to reach millions of voters in battleground states ahead of the midterms.

Read NewsGuard’s exclusive report.
https://t.co/N6XawZZTwL

[Lorenzo Arvanitis @ZoArvanitis]

A NewsGuard investigation has uncovered a new network of news sites promoting Democrats in battleground states ahead of the midterms.

The outlets publish articles online and in print, and invest in ads on Facebook to give them a far broader audience.

https://t.co/dukc31vfN1

[Jack Brewster @jbruce218]

NEW: A shadowy network of news sites is quietly pushing Democratic propaganda in battleground states ahead of the midterm elections

https://t.co/8JvfI3I2mR

[Jack Brewster @jbruce218]

New: Research shows young people are increasingly using TikTok—instead of Google—as a search engine.

NewsGuard searched for information about news topics on TikTok and found that *1 in 5* search results contained misinformation.

https://t.co/L5LrOJ168c

[Lorenzo Arvanitis @ZoArvanitis]

https://t.co/uV0Vqfn4QS