Sam Gregory

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. #ReferenceImage #Apple #SunriseandShine lnkd.in/p/grzeHYCt

Help us explore the capabilities of open-source AI models and build tools designed to support researchers, not replace them. If you bring a different perspective or experience, we’d encourage you to apply and help shape the future of open-source investigations: www.bellingcat.com/contractor-a...

www.theguardian.com/world/2026/o...

Apple's new authenticity infrastructure, Reference Image, is "a serious piece of engineering, and a 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. www.techpolicy.press/will-apples-...

Sam Gregory analyzes Apple's new Reference Image feature on the iPhone 18 Pro — "a serious piece of engineering, and a genuine response to the epistemic crisis of AI" — and the unanswered questions it raises.

If you are interested in the issue of synthetic media and deepfakes, particularly about implications for journalism and human rights evidence, this is a really interesting read from @samgregory.bsky.social.

Apple's new authenticity infrastructure, Reference Image, is "a serious piece of engineering, and a 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. www.techpolicy.press/will-apples-...

www.theguardian.com/artanddesign...

it's possible to live a life where everyone grieves its end

Not Dolly.... 💔 youtu.be/MKeulwZ3sGE?...

On Al-Jazeera on 3 deepfake threat layers: *Consumer scams *Non-consensual sexual images + nudification apps *Corrosive ambient uncertainty via political/conflict usage. The onus musn't be on us to spot, report, hide our face but on law + platform responsibility. www.youtube.com/watch?v=wp-o...

Good @indicator.media report, in tandem with @witnessorg.bsky.social on gaps in detection tools mandated by rules that came into force on Aug 2: EU AI Act, and California AI Transparency Act. They found detectors for 6 out of 13 companies, and fooled all but one. indicator.media/p/ai-generat...

Apple Reference Image: similar to ProofMode and verified capture apps since 2011 or so...Add extra sensor data + other identifiers to enrich and harden authentic image. Appears you hold off actual verification till know need it, then upload to Apple to 'confirm it' www.cultofmac.com/news/apple-r...

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. www.huffpost.com/entry/callis...

We all agree it's a bad idea... with Nathaniel Raymond and @henkvaness.bsky.social www.cbc.ca/news/world/g...

Interesting interview with @malachy.bsky.social, a pioneer of visual investigations on what's constant and what's changing in @reutersinstitute.bsky.social reutersinstitute.politics.ox.ac.uk/news/new-yor...

1/New with Afsaneh Rigot of @thedecenter.bsky.social with work from @witnessorg.bsky.social for @carnegieendowment.org: "Generative AI and Atrocity Denial in War." How AI has turned doubt itself into a weapon, from Iran to Gaza to Sudan. Thread: carnegieendowment.org/research/202...

Can AI deepfake detectors always reliably prove whether an image is real or not?

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.bsky.social analysis on where the EU AI Act gets it right, and where we need more. www.witness.org/witness-says...

In @techpolicypress.bsky.social shirin anlen and I on a foreseeable fiasco: how Google briefly turned 1 of 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 www.techpolicy.press/google-earth...

What "on earth"? The prioritization of ground truth for journalism, and our shared sense of reality - is less important to Google than "custom images to help you visualize history, create real estate plans, and more." And no, us checking every image is not the answer. blog.google/products-and...

🎥 Deepfakes: Can you trust anything you see online anymore? Episode 3 in @tbij.bsky.social series 𝑴𝒊𝒔𝒂𝒍𝒊𝒈𝒏𝒆𝒅: 𝑾𝒉𝒆𝒏 𝑨𝑰 𝑮𝒐𝒆𝒔 𝑾𝒓𝒐𝒏𝒈 out now! I talk the reasons we need to worry, why we need to act, and winners/losers in the deepfakes fight (spoiler: not ordinary folks) www.youtube.com/watch?v=_s0x...

😢https://www.theguardian.com/football/2026/jul/20/kevin-keegan-liverpool-legend-and-former-newcastle-manager-dies-aged-75?CMP=Share_AndroidApp_Other

Sigh ... www.lancaster.ac.uk/news/ai-face...

AI is eroding our shared sense of reality - that's the clear message in the Independent International Scientific Panel report. So what do we do about it? In my intervention at the UN Global Dialogue on AI, I highlighted key steps + the urgency of collective action. www.linkedin.com/posts/samgre...

humanrightsdefenders.blog/2026/06/15/i...

A must-read; could literally save someone's life

Bit of doomscrolling relief for you here. Genuinely sweet story. www.theguardian.com/lifeandstyle...

The longer you look at this, the worse it gets

RIP Barney Frank www.nytimes.com/2026/05/20/u...

So maddening to see - the devastating impact of cuts on the independent and exile media of Burma/Myanmar, just when they are needed most. reutersinstitute.politics.ox.ac.uk/news/running...

www.theguardian.com/tv-and-radio...

"When WIRED called the Zambian embassy in Washington, a member of the staff answered the phone and transferred the call to another staff member who then picked up for several seconds before hanging up. A follow-up call went unanswered."

Following our April 29 announcement, we believe it's important to be transparent about the context that led to the decision. We want to explain, where we can, why this was made on such short notice, only days before we were set to host RightsCon. Read: www.rightscon.org/rc26-stateme...

Some useful public speaking tips from Freud here.

I wrote about RightsCon's cancellation after Zambia's government announced it would postpone the event to sort out its concerns over "thematic issues." With thousands about to depart (or already en route) to Lusaka, it's a huge blow to the digital and human rights field.

This is what Marco Rubio decided to do with his life: "AIDS Creeps Back in Parts of Zambia, a Year After U.S. Cuts to HIV Assistance: A once-robust HIV treatment and prevention system, credited with saving hundreds of thousands of lives, has begun to crumble..." www.nytimes.com/2026/04/25/h...

Great conversation with @tainalon.bsky.social and @felixsimon.bsky.social 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 www.youtube.com/watch?v=yYwu...

📌 As generative AI blurs the line between real and fake, fears of a collapse in trust are everywhere. Yet audiences still believe some sources. This panel explores what has really changed in the age of synthetic media — and who is responsible for protecting truth.

“We don’t need to give up on the ability to discern what is real from synthetic,” he said. “But we do need to act fast.” In @theguardian.com I pointed to the tech companies failure to act, as Meta's own Oversight Board just also said about C2PA implementation www.theguardian.com/technology/2...

Sora “normalized a world in which people are really uncertain about what they are seeing in their timelines both when it matters & when it doesn’t matter.. whether it’s casual comedy…or conflict footage from Iran. The consequences will reverberate even if the app is gone” The Verge archive.ph/AqRO7

Sora “normalized a world in which in people are really uncertain about what they are seeing in their timelines both when it matters & when it doesn’t matter… whether it’s casual comedy… or conflict footage from Iran. The consequences will reverberate even if the app is gone.” @samgregory.bsky.social

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, and shameful normalization of likeness theft. Thanks, but no thanks, Sam www.linkedin.com/feed/update/...

Rule 1: Never trust a good foot pic. It could be AI.

An AI avatar as the apotheosis of fantasy (and driver to monetized synthetic foot pics). I talked to @drewharwell.com at the Washington Post about Jessica, the MAGA dream girl.

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 Archis Chowdhury @ Boom Live on implications of Netanyahu AI or not vids www.boomlive.in/explainers/n...