The Aporia · Nvidia’s case for taming AI agents

Nvidia’s VP of agentic AI, Adel El Hallak, argues that controlling how AI agents operate through their "harness" is crucial for AI safety, rather than slowing down model development. He claims that technical safeguards can secure AI without creating fear and uncertainty, contrasting this with calls from figures like Anthropic CEO Dario Amodei to slow progress due to rapid advancement risks. While El Hallak supports industry-led initiatives like the Open Secure AI Alliance, Democratic Representative Pete Aguilar advocates for federal regulation, arguing tech companies cannot be trusted to regulate themselves adequately.

Written locally by qwen2.5:14b on 2026-09-18, using this article's own text rather than the other coverage of the same event (that is the story summary below).

Story summary

Nvidia's Vice President of Agentic AI, Adel El Hallak, argued against fears surrounding artificial intelligence (AI) during an interview with Semafor. He emphasized that the systems supporting AI models are crucial for their safety and dismissed concerns about fear, uncertainty, and doubt (FUD). In contrast to those calling for a slowdown in AI development due to safety risks, Nvidia CEO Jensen Huang supports continued rapid progress while maintaining safety measures. El Hallak noted that advancements in AI models allow for both speed and security, countering the notion that companies cannot effectively safeguard advanced technology on their own. Rep. Pete Aguilar countered this viewpoint, advocating for federal regulations instead of relying solely on industry self-regulation.

Written for “AI Regulation Discussion” on 2026-09-18, grounded in this article and the 0 other(s) covering the same event.

Why this leaning score

The model judged this article politically coded and scored it +0.65, but all 4 of its quote(s) are attributed speech - words the article quotes from someone, not the article's own narration, so the score is not published.

Written under an earlier scoring contract, which gave a paragraph rather than checkable quotes. Re-analysing this article replaces it.

Leaning score withheld for article 17111: attributed speech only · logged 2026-09-18

📰 AI Regulation Discussion

Technology · 1 article(s) covering the same event.

Article leaning vs. publisher reliability

Source leaning vs. consistency

This article reads unscored and hedges 25% of its claims. Each row says how that neighbour differs.

September 13, 2026 different event · 95%

Letters from an American

⚖️ Leans left 🔴 15% hedged 10 of 65

“The articles discuss different perspectives on AI safety and regulation, with Article A focusing on Dario Amodei's essay and Article B highlighting Nvidia’s VP of agentic AI's viewpoint.”

Tyler Cowen: The AI Slowdown Is Not What It Seems different event · 90%

The Free Press

⚖️ Leans left 🔴 8% hedged 1 of 12 📰 publisher trust 96

“The articles discuss different perspectives on AI safety and regulation, but they are not reporting on the same specific incident or occurrence.”

Andrew Yang joins chorus of AI skeptics, warns it’s time to tap brakes while sharing juicy gossip: ‘The fear is real’ different event · 90%

New York Post

⚖️ leaning not scored 🔴 28% hedged 7 of 25 📰 publisher trust 59

“The articles discuss different perspectives on AI regulation and safety from two separate individuals (Andrew Yang and Adel El Hallak) with distinct viewpoints.”

Semafor · 392 article(s) · 0 correction(s) detected

No corrections detected for this publisher. That may mean careful reporting, or simply that nothing has been checked.

J.D. Capelouto

10 article(s) here · 1 carrying a prediction

🔮 El Hallak’s argument provides insight into the more technical case — as opposed to the geopolitical or ideological ones — for why AI doomerism may be overblown.

2026-09-17 · mixed framing · Nvidia’s case for taming AI agents

🔮 Over the summer, a hacker was able to deploy frontier AI models to breach a European IT and software company in less than 10 hours — something that would ordinarily take human hackers around two weeks, said Palo Alto Networks, which helped defend the firm against an extortion attempt.

2026-09-16 · assertive framing · Anatomy of an AI-powered hack

🔮 The attack tool, known as a worm, could allow someone to hijack WeChat accounts and spread automatically through victims’ contacts.

2026-09-15 · mixed framing · US firm alerts WeChat to AI hacking tool

🔮 The plans require rethinking safety “from the ground up,” CEO Marc Theermann told Semafor, since he wants to make robots that people will be comfortable hugging.

2026-09-15 · assertive framing · New startup looks to build robots for entertainment

🔮 Unlike in the US, where many workplaces are banning the glasses, Chinese companies that allow them face other security risks since they could in theory amass businesses’ data, including sensitive documents or internal codebases.

2026-09-15 · assertive framing · China makes enterprise push for AI glasses

🔮 In absence of that, the private sector will have no choice but to lean on unofficial cross-border communication and collaboration to keep the next WeChat worm from leaving the lab.

2026-09-15 · assertive framing · Why we can't rely on governments to handle AI safety

🔮 More efficient power conversion reduces the amount of electricity that data centers would otherwise lose as heat, which requires more water to cool servers.

2026-09-12 · assertive framing · Chipmaker says data centers should report energy waste

🔮 It’s also a partner on a privately funded, AI-planned project that aims to send a craft to Alpha Centauri — the nearest star system to Earth — a trip that would take an estimated 70,000 to 75,000 years.

2026-09-06 · assertive framing · Startup launches AI physics lab to maximize data center efficiency

More on this subject from J.D. Capelouto

Why we can't rely on governments to handle AI safety

2026-09-15 · Semafor · 63% similar

Anatomy of an AI-powered hack

2026-09-16 · Semafor · 59% similar

US firm alerts WeChat to AI hacking tool

2026-09-15 · Semafor · 58% similar

AI deployment in businesses outpaces trust, study finds

2026-09-06 · Semafor · 57% similar

All 10 articles by J.D. Capelouto →

Anthropic News Nvidia Semafor the Open Secure AI Alliance

Nvidia ORG · 2× Adel El Hallak PERSON · 1× Anthropic ORG · 1× Dario Amodei PERSON · 1× El Hallak PERSON · 1× El Hallak’s PERSON · 1× Huang PERSON · 1× Jensen Huang PERSON · 1× News ORG · 1× Semafor ORG · 1×

“The world’s not going to slow down.” Room for Disagreement Some in Washington argue tech companies can’t be trusted to take up these issues on their own. “This is not an industry that can regulate itself,” Democratic Rep. Pete Aguilar said this week, arguing for the need for federal industry-wide rules on AI safety.

framing: mixed · carried by 1 article(s) · first seen 2026-09-18

🔮 El Hallak’s argument provides insight into the more technical case — as opposed to the geopolitical or ideological ones — for why AI doomerism may be overblown.

2026-09-18 · Semafor

Nvidia’s case for taming AI agents · mixed framing

The scaffolding that’s built around an AI model is just as important — if not more important — for AI safety than the model itself, Nvidia’s VP of agentic AI told Semafor. asserted

VP → build → Semafor

“Rather than creating fear and FUD, let’s just secure it. asserted

’s → create → it

We know how. asserted

We → know → ?

We’ve done these things before,” Adel El Hallak said, referencing the acronym for fear, uncertainty and doubt. asserted

Hallak → do → fear

El Hallak’s argument provides insight into the more technical case — as opposed to the geopolitical or ideological ones — for why AI doomerism may be overblown. uncertain

doomerism → provide → ones

Know More Nvidia CEO Jensen Huang has emerged as the most prominent tech figure voicing skepticism about recent claims by leaders like Anthropic CEO Dario Amodei that AI labs should slow down model development because the tech is quickly advancing beyond companies’ abilities to safeguard it. uncertain

tech → know → it

AI agents are central to that fear because they can take actions on their own, though regulation of agentic AI specifically is not discussed as much as is broad oversight of the models. asserted

oversight → take → models

Huang this week rejected the notion that the industry has to choose between safety and speed, saying “you could definitely have both at the same time.” uncertain

you → reject → time

El Hallak said advances in AI models make “harness engineering” more critical — a harness is the layer of instructions and code that governs how an AI agent operates in practice. asserted

agent → say → practice

“If you’re training a capable agent, it is important that you don’t put a lion in and expect to be able to control it in a horse’s pen.” asserted

you → train → pen

He couldn’t comment on whether government regulation of the agentic layer is needed, but pointed to industry initiatives like the Open Secure AI Alliance as a venue for that kind of work, to figure out “at which part of the layer do we need to add more strengths and more guardrails and more safeguards.” asserted

we → comment → strengths

“The world’s not going to slow down.” Room for Disagreement Some in Washington argue tech companies can’t be trusted to take up these issues on their own. “This is not an industry that can regulate itself,” Democratic Rep. Pete Aguilar said this week, arguing for the need for federal industry-wide rules on AI safety. asserted

Aguilar → go → safety

Nvidia’s case for taming AI agents

Quick Summary

AI analysis runs on qwen2.5:14b, locally

Signals How these are calculated →

Story

How this is being covered How these are calculated →

Compared with similar articles

Publisher

Who wrote this

Topics

Subjects

Narrative

Claims (12 extracted, 3 hedged)