English
Sam Gregory
Report

Sam Gregory

@SamGregory

AI, trust, deepfakes and epistemic resilience. 'Prepare, Don't Panic' initiative on media manipulation/generative AI. TED speaker deepfakes. Former ED WITNESS.

11.8K Following11.7K Followers32.9K Posts
  1. Sam Gregory@SamGregory
    With @cward1e, talked to @hari for @AmanpourCoPBS deep-dive on how #AI content is impacting elections Satire to deception to resurrection to dismissing the real....highlighting detection challenges, avoiding blaming ourselves, and pipeline responsibility
    45314.5K
  2. Sam Gregory@SamGregory
    I am very excited about this.
    001693
  3. Sam Gregory@SamGregory
    Apple's new authenticity infrastructure, Reference Image, is "a serious piece of engineering, and genuine response to the epistemic crisis of AI", but a start not an end-point. I reflect on lessons from the field, unanswered questions and next-step needs.
    Download
    001198
  4. Sam Gregory@SamGregory
    Apple enters the defending reality fray with Reference Image, a digital negative to prove a real photo. I wrote about seeing another major company create the Proof Mode I've wanted for years...what I like, the questions, the concerns. #SunriseandShine
    2181.3K
  5. Sam Gregory@SamGregory
    Sam Gregory
    RT @MetacogniShane: One of the red flags of AI-written text is the heavy use of these odd linguistic constructions called “cataphoric teasers” to create an artificial feeling of suspense. These are easy to spot because they usually take the form of phrases like, “Here’s the part that nobody tells you…” or “Here’s what most people get wrong…” or “Here’s where it gets interesting.” From my experience, text written by Claude is often saturated with these types of phrases. I don't see it as much in ChatGPT (that's anecdotal, though. I might be wrong).
    Jan Zilinsky @janzilinsky
    I also thought @paulbloomatyale’s reflections about this were good: AI writing often has a clickbait-like urgency, as if every sentence was trying to keep you reading
    0181015
  6. Jan Zilinsky✓@janzilinsky
    I also thought @paulbloomatyale’s reflections about this were good: AI writing often has a clickbait-like urgency, as if every sentence was trying to keep you reading
    515244197K
  8. Sam Gregory@SamGregory
    On Al-Jazeera on 3 deepfake threat layers: *Consumer scams *Non-consensual sexual images + nudification *Corrosive ambient uncertainty via political/conflict usage. Onus can't be on us to spot, report, hide our face but on law + platform responsibility
    016501
  9. Sam Gregory@SamGregory
    Callista Gingrich is Callista Gingrich is my conclusion about whether her ambassadorial videos are AI-generated.... no, she just is like this, and is pretty consistent about that. We're in a world where she now looks like a bad corporate HR AI avatar.
    100372
  11. Sam Gregory@SamGregory
    It's August 2! It's August 2! Why am I so excited (and trepidacious)... It's the day both California and EU rules on content provenance and AI go into effect. Read on for @witnessorg analysis on where the EU AI Act gets it right, and where we need more.
    036477
  12. Sam Gregory@SamGregory
    Sam Gregory
    RT @N_Waters89: "We take misinformation seriously... [so we decided destroy trust in a key reference point in the epistemology of journalists, fact checkers and researchers]" Well done, you put Synth ID in it. Now fact-checkers can write an article after it's used to create a viral fake story.
    News from Google @NewsFromGoogle
    We take misinformation seriously – every image created with Nano Banana in Google Earth includes the SynthID digital watermark, so if someone is unsure about an image, they can ask the Gemini app or use Lens in Search to see if the image was AI-generated. In addition, we prevent image creation on harmful topics and are continually updating our protections →
    0159012
  13. News from Google✓@NewsFromGoogle
    Posts @Shayan86
    We take misinformation seriously – every image created with Nano Banana in Google Earth includes the SynthID digital watermark, so if someone is unsure about an image, they can ask the Gemini app or use Lens in Search to see if the image was AI-generated. In addition, we prevent image creation on harmful topics and are continually updating our protections →
    82339125K
  14. Sam Gregory@SamGregory
    In @techpolicypress shirin anlen and I on a foreseeable fiasco: how Google briefly turned one of the world's most trusted sources of evidence into a generator of plausible fakes, and why tech cos must consult civil society experts before launch, not after
    Download
    021442
  15. Sam Gregory@SamGregory
    Why would Google do this? Because the prioritization of ground truth for journalism, and our shared sense of reality - is less important than "custom images to help you visualize history, create real estate plans, and more." And no, us checking every image is not the answer.
    𝚑𝚎𝚗𝚔 𝚟𝚊𝚗 𝚎𝚜𝚜 @henkvaness
    Tonight I typed just one sentence into Google Earth and put refugees near the Mexican border. Then I planted a nuclear plant in Iran. What on earth is Google doing? Check my latest post here:
    2210952
  16. 𝚑𝚎𝚗𝚔 𝚟𝚊𝚗 𝚎𝚜𝚜✓@henkvaness
    Tonight I typed just one sentence into Google Earth and put refugees near the Mexican border. Then I planted a nuclear plant in Iran. What on earth is Google doing? Check my latest post here:
    Download
    1601.2K7.4K3.6M
  17. Sam Gregory@SamGregory
    🎥 Deepfakes: Can you trust anything you see online anymore? New @TBIJ series 𝑴𝒊𝒔𝒂𝒍𝒊𝒈𝒏𝒆𝒅: 𝑾𝒉𝒆𝒏 𝑨𝑰 𝑮𝒐𝒆𝒔 𝑾𝒓𝒐𝒏𝒈 out now! The reasons we need to worry, why we need to act, and winners/losers in the deepfakes fight (spoiler: not us)
    011467
  19. Sam Gregory@SamGregory
    Sam Gregory
    RT @mediapartyorg: How can journalists navigate the rising threat of deepfakes and synthetic media? @SamGregory introduces the Deepfakes Rapid Response Force, a project from @witnessorg that connects reporters with forensic experts to verify suspicious content and protect reporting integrity. Watch the full Lightning Talk from #MediaParty 2025 now available here👇
    030503
  20. Sam Gregory@SamGregory
    Great conversation with Tai Nalon and Felix Simon at #ijf26 on "When seeing is no longer believing: perils and possibilities in a world of synthetic media" with our differing viewpoints on scope, societal implications, and what to do
    002319
  21. Sam Gregory@SamGregory
    OpenAI is shutting down Sora. But the damage is done: normalized AI slop, eroded trust in what's real, detection inadequate, provenance tools still MIA without senior company leadership, shameful normalization of likeness theft. Thanks, but no thanks, Sam
    002178
  22. Sam Gregory@SamGregory
    Sam Gregory
    RT @mahsaalimardani: Iran is the worst-case scenario for everything we’ve been warning about: AI fakes, civilian footage dismissed as fabricated, and underneath it all, a regime that has engineered the doubt itself. “The problem is curated access. The regime controls who gets to document and what gets documented.” I spoke to @morganmeaker for Bloomberg Weekend on the information war in the Iran war.
    0601.4K
  23. Sam Gregory@SamGregory
    Hyper-realistic AI both to create fantasy images that illustrate people's hopes, and where the wish that something is fake finds a ready-made access point via poor forensic analysis: Talked with @chowarchis @boomlive_in about is it/isn't it Netanyahu vids
    012249
  24. Sam Gregory@SamGregory
    Sam Gregory
    RT @mahsaalimardani: New piece with shirin anlen based on our work at @witnessorg on how 'forensic cosplay' (h/t @CraigSilverman) is being used to discredit authentic photos of civilian casualties in the Iran war. The fake heatmap and "ELA" case spread through my own networks. People I know concluded the photos were fake, primed by justified distrust of state sources. That distrust is legitimate. But it can lead to dangerous denialism. With references to @Shayan86 and @talhagin who have been working on the deluge of AI cases during this war and insights from forensics specialists.
    01306.1K
  25. Show more posts