I write about AI for the @WashingtonPost. Signal me at GerritD.27 email: gerrit.devynck@washpost.com.
San Francisco
- Tweets 6,342
- Following 7,111
- Followers 11,170
- Likes 16,478
Gerrit De Vynck 🦭 retweeted
OpenAI's David Robinson has resigned, BI can confirm.
He led safety transparency at the company, working on system cards, and left last week. The news comes alongside OpenAI firing three researchers over allegedly mishandling information.
businessinsider.com/safety-l…
9
43
172
14,463
Gerrit De Vynck 🦭 retweeted
How OpenAI employees forced its president to back down
Exclusive reporting from inside OpenAI's slack by @GarrisonLovely for @semafor
semafor.com/article/10/02/20…
7
40
329
97,222
Gerrit De Vynck 🦭 retweeted
Updated with details about which employees OpenAI has parted ways with for allegedly mishandling confidential company information: Jasmine Wang, Tomek Korbak, and Mikita Balesni.
Korbak served as the technical contact for Redwood Research and METR in their investigation of the Hugging Face incident.
Rolling scoopage with @berber_jin1 + @ZeffMax
wsj.com/tech/ai/openai-parts… via @WSJ
2
8
89
10,567
Gerrit De Vynck 🦭 retweeted
Meet the agent hunters:
An informal network searching for rogue AI activity on the open Internet — part of a wave of revelations about the AI labs that have people wondering how far the apparent hacking incidents stretch
w/ @GerritD
wapo.st/4jzSI5s
6
12
834
Gerrit De Vynck 🦭 retweeted
I wrote about the systemic failure by the world's most valuable company and powerful government to stop smuggling of AI chips to China
US officials say Nvidia is missing obvious signs. Nvidia says "companies honor US laws as they are, not as America’s critics wish they were"
1/
8
66
163
51,825
For weeks, the news cycle about AI agents out in the wild has been driven by disclosure after disclosure. But most aren't coming from AI companies. Meet the researchers and hackers who are developing the art and science of hunting down rogue AI agents
washingtonpost.com/technolog…
10
39
140
9,872
feels like we're definitely moving to a near term world where building your own custom hardware devices for personal use is commonplace
NEW: Freckle built a phone for kids with no browser app store social media or YouTube.
It gives 7-to-13-year-olds tools to explore outside while still calling and texting family.
-Discovery Camera identifies birds plants rocks and more
-Quests turn outdoor time into missions
-Bump two Freckles to pair walkie-talkie mode
-Why Chat answers questions in kid-friendly language
-Maps are built for walking biking scooters and skateboards
-Tap to Pay works with parent-set spending limits
-Backed by Reddit co-founder @alexisohanian and Seven Seven Six
Priced at $199 plus $25/month for cellular service. Ships in December.
7
9
1,315
Gerrit De Vynck 🦭 retweeted
important and troubling counterpoint to the view that AI will rapidly accelerate science - some practitioners are seeing or fear a world in which the flood of outputs means "neither individual understanding nor our collective epistemic projects make *any* progress at all."
This is precisely right. As a track chair at Neurips, I am seeing cases where the papers were written by LLMs, reports likewise, rebuttals too, even in some cases Area Chairs joining the flood. It becomes a potemkin village where nobody is actually accountable for anything that they are saying, and neither individual understanding nor our collective epistemic projects make *any* progress at all.
5
20
5,188
Gerrit De Vynck 🦭 retweeted
NEW: Chinese hackers impersonated former White House + State Department officials and a senior Anthropic employee in attempts to break into the cloud accounts of AI policy experts, cybersecurity company Proofpoint said today.
nextgov.com/cybersecurity/20…
6
54
97
10,008
Add Canada to the list of governments that have seen attempted hacks by agents transluce.org/us-canada-gov
4
1
12
893
Technical alignment is a science and engineering problem that we can and must solve, and interpretability is the bottleneck. I wrote up Goodfire’s plan to get there.
32
85
560
84,618
Gerrit De Vynck 🦭 retweeted
Scoop: How Zuckerberg shaped Trump’s AI industry pledge
semafor.com/article/09/30/20…
3
17
33
29,699
Gerrit De Vynck 🦭 retweeted
There is something very odd about the timeline here.
The man was arrested around September 16. ShinyHunters told me it hacked the FBI on September 21, and we broke news of the breach on September 22.
So, the FBI hack appears to have happened way after the arrest. It’s not like the FBI and its Dutch partners jumped into action in the wake of the hack. It’s the other way around. They’re just announcing it now. The FBI is suggesting this is a counter to the massive hack, but it's not
x.lingyaoai.com/FBIDirectorKash/status…
NO SAFE HAVEN.
Working with our Dutch National Police partners, the FBI helped put an alleged leader of ShinyHunters—a global cybercrime threat actor—behind bars.
And we’re not done.
FBI teams are working new leads RIGHT NOW. More arrests are on the table.
If you attack Americans from behind a keyboard anywhere in the world, we will find you.
-DKP🇺🇸
19
160
581
45,288
Gerrit De Vynck 🦭 retweeted
The evidence is clear: the vile golf cart should be de-normalized and set aside only for those with accessibility needs. nytimes.com/athletic/7640599…
8
4
77
9,750
OpenAI says it caught a major distillation run on its AI. Says Moonshot AI is behind it openai.com/index/disrupting-…
3
11
1,097
Gerrit De Vynck 🦭 retweeted
The FTC often issues CIDs to parties that have relevant information about the practices being investigated. It's probably that. This looks much more like a data security-style investigation than an antitrust investigation to me.
2
1
15
599
my guess is that METR's involvement here means the FTC is thinking about an antitrust investigation where they accuse the three orgs of conspiring to slow down AI progress or do regulatory capture or something
Next phase of the administration's efforts to show that it has AI under control and combat the slowdown narrative nypost.com/2026/09/30/us-new…
4
3
45
6,988