Sign in

Patrick Grady

@patrickgradyy.bsky.social
297 followers 304 following 235 posts

🇪🇺 Head of European Tech @itif.org

PostsRepliesMedia
Patrick Grady @patrickgradyy.bsky.social · 22h
It may be convenient to pretend that agents are dressed up and walk around on their own. But it is an illusion. Like the emperor in Hans Christian Andersen’s fable, AI agents have no clothes. halffullai.substack.com/p/ai-agents...
halffullai.substack.com
AI Agents Have No Clothes
Blaming software for company mishaps creates a moral hazard
010
Patrick Grady @patrickgradyy.bsky.social · 01/10/2026
Important and serious work alignmentalignment.ai/caaac
alignmentalignment.ai
Center for the Alignment of AI Alignment Centers
We align the aligners
010
Patrick Grady @patrickgradyy.bsky.social · 24/09/2026
www.politico.eu/article/don...
politico.eu
Don’t believe the doomers: Europe’s tech industry slams AI panic
The existential fear emanating from Silicon Valley and the U.N., techies say, masks what is ultimately a cybersecurity challenge.
001
Patrick Grady @patrickgradyy.bsky.social · 23/09/2026
Sell European is infinitely more important than Buy European www.bloomberg.com/news/articl...
bloomberg.com
ASML Executive Says Europe’s Biggest Firm Has No Sales in Europe
A top executive at ASML Holding NV, Europe’s most valuable company, said it isn’t selling any of its chipmaking machines on the continent and that the region risks getting left behind as the US, China and India invest heavily in developing their domestic semiconductor industries.
010
Patrick Grady @patrickgradyy.bsky.social · 23/09/2026
Net flow of AI talent (authors of publications and inventors behind patents) is plummeting in the US... opportunity for Europe? hai.stanford.edu/assets/file...
000
Patrick Grady @patrickgradyy.bsky.social · 20/09/2026
The EU should not "pace the frontier" ◼ Current alarm-ringing reflects the culture in American AI labs, not new risks. ◼ Labs have called for a pause before. They didn’t stop then; they won’t stop now. ◼ The EU’s AI Act will do nothing to slow things halffullai.substack.com/p/no-europe...
020
Patrick Grady @patrickgradyy.bsky.social · 18/09/2026
Europe should ignore this noise and push ahead with developing and adopting AI. In the face of this latest AI panic imported from the US, sobriety and stoicism will advantage the old continent. www.mlex.com/mlex/articl...
mlex.com
AI slowdown calls could push EU AI Act toward 'Brussels effect' | MLex | Specialist news and analysis on legal risk and regulation
AI companies' calls to slow frontier-model development could help the EU extend the AI Act's global influence, echoing the "Brussels effect" created by the GDPR. But the European Commission has been reluctant to engage on the issue and has largely stressed enforcement in its own backyard. For some policy experts, this is a missed opportunity.
020
Patrick Grady @patrickgradyy.bsky.social · 18/09/2026
... and European leaders don't actually believe these latest warnings, this victory lap is a complete illusion.
100
Patrick Grady @patrickgradyy.bsky.social · 18/09/2026
2️⃣ Others believe the AI Act will give the EU a top seat in shaping a global AI agreement because they put together the GPAI Codes. This *vastly* overstates the bloc's relevance in the eyes of these leading labs. For them, there are only two seats at the global AI table.
100
Patrick Grady @patrickgradyy.bsky.social · 18/09/2026
Yes—labs need to prove their AI is safe to use in the EU, and hand in evidence on request. But—an AI going truly and catastrophically rogue in the US will not think twice because of the EU's AI Act. None of this paperwork will matter in the face of an actual extinction event.
100
Patrick Grady @patrickgradyy.bsky.social · 18/09/2026
1️⃣ Tin foil hat on for a moment. If EU policymakers believe these latest warnings (they shouldn't), how exactly will the "systemic-risk" obligations in the AI Act prevent AI extinction?
110
Patrick Grady @patrickgradyy.bsky.social · 18/09/2026
"The EU has already legislated for exactly this scenario". The feeling in Brussels is that the AI Act has prepared the bloc to regulate world-ending AI. That's either (1) naive or (2) hubristic 🧵 www.politico.eu/article/eu-...
politico.eu
Keep calm and regulate on: Inside the EU’s response to AI extinction warnings
Brussels sees the panic set off in Silicon Valley as vindication that its AI regulation is the right path.
110
Patrick Grady @patrickgradyy.bsky.social · 16/09/2026
2024: We need to speed up AI to catch up with the US! 2025: We need to speed up AI to catch up with the US! 2026: We need to slow down because two US companies told us to 🤦
000
Patrick Grady @patrickgradyy.bsky.social · 16/09/2026
Initial red flags: ❗ Scope is huge and slippery. ❗Guardians/parents will have significant responsibility for that critical 13-14 age group. That's not without risk ❗Where's the balance? Are all these services and features bad? Where's the evidence?
022
Patrick Grady @patrickgradyy.bsky.social · 16/09/2026
✔️ Some good ideas, especially excluding educational services (Duolingo streaks live on!), tiering obligations, and relying on the EU's eID solution (which is FAR from ready, but a better solution than any verification/estimation tool currently on the market).
122
Patrick Grady @patrickgradyy.bsky.social · 16/09/2026
◼ No social media plus for under 13s. 13- to 14-year-olds can use restricted versions*, with guardians' permission ◼ *where guardians can control users' experience ◼ App stores will police the age gate; users will use the EU's eID ◼ New general safety-by-design obligations
122
Patrick Grady @patrickgradyy.bsky.social · 16/09/2026
🚨 So the EU's social media ban (EU KIDS ACT) is finally here. 🧵
165
Patrick Grady @patrickgradyy.bsky.social · 13/09/2026
We ought to think critically about AI’s impact on society before importing gloomy sci-fi narratives from the streets of San Francisco. Half Full AI will serve as an optimist’s guide to AI developments. Check it out 👇 halffullai.substack.com/p/half-full...
halffullai.substack.com
Half Full AI
An optimist’s guide to AI developments and policy
000
Patrick Grady @patrickgradyy.bsky.social · 10/09/2026
The EU could really differentiate itself by ignoring this nonsense and continuing to address actual risks / adopt AI where it makes sense. Undoubtedly, figity and unserious policymakers will fly into the light and import yet another micropanic from the streets of San Francisco.
000
Patrick Grady @patrickgradyy.bsky.social · 28/08/2026
The only people not panicking about AI cyber attacks? Cyber experts. “They were not going rogue. They were doing what humans had told them to do, with the precocity and indiscipline of talented but unsupervised children. They did no real harm.” economist.com/by-invitatio...
economist.com
Fears of AI-induced armageddon are overdone
Applying basic security principles would go a long way to fixing problems identified by Anthropic and OpenAI, writes Ciaran Martin
020
Patrick Grady @patrickgradyy.bsky.social · 26/08/2026
👏
000
Patrick Grady @patrickgradyy.bsky.social · 20/08/2026
I'm a fan of Betteridge's law that "any headline that ends in a question mark can be answered by the word no." Philosophers are sceptical of attributing minds to machines, for they cannot experience as humans do. Thinking AI is conscious is a harmful and distracting delusion.
000
Patrick Grady @patrickgradyy.bsky.social · 18/08/2026
Bluesky really has its finger on the pulse
010
Reposted by Patrick Grady
Diane Coyle @dianecoyle1859.bsky.social · 17/08/2026
My book The Measure of Progress is out soon in paperback - available for pre-order! press.princeton.edu/books/paperb...
press.princeton.edu
The Measure of Progress
Why do we use eighty-year-old metrics to understand today’s economy?
422847
Patrick Grady @patrickgradyy.bsky.social · 17/08/2026
Unsure whether the text you're reading is treating software as if it had human traits such as consciousness, beliefs, or intentions? Paste your text into the deanthropomophiser to see things more clearly 👇 deanthropomorphiser.lovable.app/
deanthropomorphiser.lovable.app
The deanthropomorphiser
Machines do not want, know or decide. Score your text and fix the wording.
010
Patrick Grady @patrickgradyy.bsky.social · 10/08/2026
Nice The Economist picked up my letter on Musk’s latest musings. No one’s guesses are gospel and silly claims distract policymakers from present challenges and are then forgotten about. That’s why I created the Archive of Incorrect AI Predictions 👉 boyswhocriedai.lovable.app
000
Patrick Grady @patrickgradyy.bsky.social · 26/06/2026
Check out the proposal and full list of responses here: ec.europa.eu/info/law/be...
ec.europa.eu
European Commission - Have your say
European Commission - Have your say
000
Patrick Grady @patrickgradyy.bsky.social · 26/06/2026
5️⃣ Monitor and communicate progress The proposal includes a 5-year reporting cycle (attached). That's too slow. Publish annual reports with KPIs, case studies, cost comparisons, and recommendations. Use it to advertise the initiative and drive uptake across Europe 📊
120
Patrick Grady @patrickgradyy.bsky.social · 26/06/2026
4️⃣ Create a fully centralised EU registry from day 1 The proposal creates a "central interface" that connects national registries. That's not enough. A true single market needs a fully centralised registry from the start, not 18 months later via implementing acts 🧩
100
Patrick Grady @patrickgradyy.bsky.social · 26/06/2026
3️⃣ Provide legal certainty throughout the company lifecycle What happens when a company restructures? Goes bankrupt? Changes valuation? Founders dispute? The framework needs clearer guidance for these scenarios. Investors need certainty before they commit capital 💶
100
Patrick Grady @patrickgradyy.bsky.social · 26/06/2026
2️⃣ Permit employment practices that reflect modern startups Remote-first, cross-border hiring that prioritises talent shouldn't trigger burdensome cross-border branch requirements (as we state below). The proposal shouldn't punish digital-first companies for scaling quickly 🚀
100
Patrick Grady @patrickgradyy.bsky.social · 26/06/2026
1️⃣ Clarify the interaction between EU Inc and national laws. Article 4 needs more clarity on which laws (tax, labour, liability, disputes) apply where, with minimum harmonisation standards. We've been calling for this since 2024; it's time to get it right 📜
100
Patrick Grady @patrickgradyy.bsky.social · 26/06/2026
🇪🇺 EU Inc could be Europe's answer to Delaware–if designed right. The EU got the setup right: 48-hour, €100, digital-only formation... But it's still complicated to scale across the bloc. Here are 5 recommendations to improve the framework 🧵 progresschamber.org/resources/s...
progresschamber.org
Submission to the Call for Evidence on The Proposed Regulation on the 28th Regime Corporate Legal Framework (‘EU Inc’) - Chamber of Progress
111
Patrick Grady @patrickgradyy.bsky.social · 26/06/2026
Check out the proposal and full list of responses here: ec.europa.eu/info/law/be...
ec.europa.eu
European Commission - Have your say
European Commission - Have your say
000
Patrick Grady @patrickgradyy.bsky.social · 26/06/2026
5️⃣ Monitor and communicate progress The proposal includes a 5-year reporting cycle (attached). That's too slow. Publish annual reports with KPIs, case studies, cost comparisons, and recommendations. Use it to advertise the initiative and drive uptake across Europe 📊
120
Patrick Grady @patrickgradyy.bsky.social · 26/06/2026
4️⃣ Create a fully centralised EU registry from day 1 The proposal creates a "central interface" that connects national registries. That's not enough. A true single market needs a fully centralised registry from the start, not 18 months later via implementing acts 🧩
100
Patrick Grady @patrickgradyy.bsky.social · 26/06/2026
3️⃣ Provide legal certainty throughout the company lifecycle What happens when a company restructures? Goes bankrupt? Changes valuation? Founders dispute? The framework needs clearer guidance for these scenarios. Investors need certainty before they commit capital 💶
100
Patrick Grady @patrickgradyy.bsky.social · 26/06/2026
2️⃣ Permit employment practices that reflect modern startups Remote-first, cross-border hiring that prioritises talent shouldn't trigger burdensome cross-border branch requirements (as we state below). The proposal shouldn't punish digital-first companies for scaling quickly 🚀
110
Patrick Grady @patrickgradyy.bsky.social · 26/06/2026
1️⃣ Clarify the interaction between EU Inc and national laws. Article 4 needs more clarity on which laws (tax, labour, liability, disputes) apply where, with minimum harmonisation standards. We've been calling for this since 2024; it's time to get it right 📜
100
Patrick Grady @patrickgradyy.bsky.social · 18/06/2026
BONUS: 3 misconceptions about driverless cars –  1. We will own them. In the short term, fleets of robotaxis will lead the self-driving revolution. 2. They are futuristic. They will be in Zagreb/London/Madrid/Munich this year 3. They are unsafe. Driverless taxis are ~10x safer
000
Patrick Grady @patrickgradyy.bsky.social · 18/06/2026
Self-driving cars will provide a welcome boost to Europe's economy, green ambitions, and improve road safety. Read more about how here: progresschamber.org/insights/eu...
progresschamber.org
Europe's Self-Driving Future Is Within Reach
Robotaxis arrive in Europe in 2026. New research shows self-driving cars could add €9B to the EU economy and prevent thousands of crashes—if policymakers act.
100
Patrick Grady @patrickgradyy.bsky.social · 18/06/2026
🚗 Lift fleet number limits —> there ought to be no delay. 🚦 Update the road rules —> laws to include driverless cars, manage new licences, and clarify legal responsibilities. 👥 Win the public over! —> politicians and deployers need to inform the public about their safety.
100
Patrick Grady @patrickgradyy.bsky.social · 18/06/2026
That would: 💶 Add €9 billion to the EU’s economy, 🌱 Reduce emissions by 900 million kg CO2e (~ 1.4 million flights from Helsinki to Munich) ⚠️ Prevent up to 6,875 traffic collisions – and countless tragedies – each year. For this "optimistic scenario", we need the EU to..
100
Patrick Grady @patrickgradyy.bsky.social · 18/06/2026
🚨 NEW: Research for the Chamber of Progress reveals that by 2035, up to 15% of taxi journeys could be self-driven if we play our cards right 🧵 progresschamber.org/research/im...
progresschamber.org
Impacts of Automated Ridehailing Services in the EU: - Chamber of Progress
100
Patrick Grady @patrickgradyy.bsky.social · 16/06/2026
Social media bans are not evidence-based policymaking but populist politics. US author Jonathan Haidt met Macron in 2025 and says he convinced the French leader to act on social media bans – something Macron all but confirmed in January this year. What we've seen since, well
023
Patrick Grady @patrickgradyy.bsky.social · 12/06/2026
Check out the full Code here digital-strategy.ec.europa.eu/en/policies...
digital-strategy.ec.europa.eu
Code of Practice on Transparency of AI-Generated Content
This code of practice supports compliance with the AI Act transparency obligations related to marking and labelling of AI-generated content.
000
Patrick Grady @patrickgradyy.bsky.social · 12/06/2026
This Code is well overdue, and the Guidelines (expected in July), and the technical standards. Why rush and panic the industry? We argue policymakers should slow down: progresschamber.org/insights/ai...
progresschamber.org
On the AI Act, the EU Need Not Move Fast and Break Things: - Chamber of Progress
Policymakers should slow down the rollout of the AI Act, not to water it down, but to ensure its proper enforcement
100
Patrick Grady @patrickgradyy.bsky.social · 12/06/2026
❌ The bad: - 1.2 puts obligations on upstream providers that go well beyond the scope of Article 50 - 2.1.1 provides flexibility but creates an overly expansive obligation for signatories to share detection solutions - 4.2 this premarket testing is not required by the AIA!
100
Patrick Grady @patrickgradyy.bsky.social · 12/06/2026
*as pointed out in our response, this technology is unproven at scale and technically infeasible: progresschamber.org/resources/r...
progresschamber.org
Response to EC Survey on AI Act Transparency Code of Practice 2nd Draft (30 Mar) - Chamber of Progress
100
Patrick Grady @patrickgradyy.bsky.social · 12/06/2026
✅ First, the good: - Measure 1.1.2 no longer requires adding imperceptible watermarks to audio, image, or video context* - 1.3 & 1.4 are optional measures, supporting the AI Act's distinction between providers and deployers - 2.4 provides flexible guidance around AI literacy
100