Sentinel

Our elite cyborg forecasters identify and discuss the most important warning signs of global catastrophic risks for you.

[Thomas Larsen @thlarsen]

We at AIFP have made a big update towards shorter timelines: all of our medians for superhuman AI researcher (SAR) are now between May and Nov 2027.

That being said: the state of public evidence on AI timelines is inadequate. Most of the evidence from the recent METR report was redacted. So we're reading the tea leaves.

AI companies should (1) release way more evidence about timelines and (2) actually start pacing the frontier, especially if the timelines are as short as we think.

[Daniel Kokotajlo @DKokotajlo]

Unfortunately, it seems like the intelligence explosion will be happening in 2027 after all. Most of this post analyzes the new METR report (though other sources of evidence are also important). If they’re right, that implies an AI-driven acceleration that is 5 months ahead of the AI 2027 scenario. I am scared, sad, and angry. https://t.co/7KWmdc6AZk

[Andrew Curran @AndrewCurran_]

The AI Futures Project have updated their timelines. They say we are five months ahead of AI 2027, and they now believe a Superhuman AI Researcher will arrive sometime in 2027.

[Sentinel @XriskFYI]

@LPacchiardi Thank you!

[MTS @MTSlive]

SITUATION EXPLAINED: Forecasters now put superintelligence at early 2028, and give AI a 6% chance of wiping out humanity in the next 10 years.

• Sentinel's forecasters give a 10% chance of superintelligence by early 2027 and 90% by mid-2047
• METR estimated last week that Anthropic is improving about twice as fast as a year ago
• Daniel Kokotajlo said that pace would bring superintelligence sooner than his team's AI 2027 scenario predicted
• Individual forecasters put the chance of a US-ordered halt anywhere from 2% to 25%
• GPT-6 Astra finished about 10% of tasks labeled safety research vs 16% of the same tasks labeled capabilities research, and OpenAI's monitors didn't flag it
• Anthropic's IPO filing says it "may imminently meet" its own threshold for "dramatically accelerated progress attributable to AI"

@theojaffee: "If Anthropic really believes that we're going to get real superintelligence within the next year or two, then they're not doing nearly enough to inform the government and try to pace the frontier."

[Peter Wildeford🇺🇸🚀 @peterwildeford]

In light of recent work by METR on AI acceleration, it's fascinating that...

- Anthropic's AI R&D evals are saturated, so they jsut stopped running them
- CoBench has been was pulled for fixes
- Internal acceleration metrics are "only partially published."

So Anthropic deciding they have not hit their AI R&D threshold is now a "holistic judgment."

This threshold is 2x. Anthropic currently estimates 1.88x. 32 of their 100 refits hit 2x or more.

When Opus 5.5 was launched, the 'one-time jump' hypothesis explanation won 99/100 refits. This time it now wins in 0/100 refits. So looks like it's not a one-time jump.

Anthropic has much better internal evidence they are not sharing. And they have an internal model that is stronger than Fable 5.2 that they are not sharing.

[Sentinel @XriskFYI]

Brent crude oil fell below $100 per barrel after the US elections, reflecting increased hope that the Iran war will end sooner rather than later. The elections, on top of some level of insecurity about the AI trade, caused some turbulence in the stock market as well. And, the US wants China to increase the supply of rare earth minerals.
7/7

[Sentinel @XriskFYI]

The November 9 Sentinel Report is out!
Global Risks Weekly Roundup #45/2026
Anthropic warns of imminent dramatically accelerated progress due to AI, METR echoes their warning, US midterms proceed peacefully
1/7
https://t.co/ticmIz0SWI

[Sentinel @XriskFYI]

The Iran war remains at a stalemate. Trump could still potentially order a new round of strikes. In Israel, Eisenkot will try to form a new government. European authorities are investigating whether Russia is behind recent hoax bomb threats to air travel.
6/7

[Nuño Sempere (in Asunción) @NunoSempere]

Ayayai, forecasters wanted to be at red, I wanted to be at yellow, we split the difference and Sentinel now has the highest alert level in a long time now.

[Sentinel @XriskFYI]

This week, we forecasted about
• cyber risks to banks and the power grid
• the risk of a $1B+ cyberattack
• the chance of an agreement between the US and Iran by the end of the year
• how high Brent crude might go before the end of the year
2/2

[Sentinel @XriskFYI]

The October 26 Sentinel Report is out!
Global Risks Weekly Roundup #43/2026
Critical cyber vulnerabilities sparked a wave of cyberattacks, OpenAI agents weakened monitoring systems, Brent crude hit $117
1/2
https://t.co/uTnjO482sR

[Nuño Sempere (in Asunción) @NunoSempere]

https://t.co/IJriH1THUp

[Sentinel @XriskFYI]

This week, we forecasted about the chances of
• a $1B+ AI-enabled cyberattack by 2028
• a severe rogue AI incident before 2028
• Russian diesel exports this year
• a return of shipping traffic in the Strait of Hormuz this year
• Ebola spreading to another country
2/2

[Sentinel @XriskFYI]

The October 12 Sentinel Report is out!
Global Risks Weekly Roundup #41/2026
Hundreds of mathematical solutions found by OpenAI, US and Russia reach diesel export deal, Houthis attack Saudi airport
1/2
https://t.co/OT3pVDzZzx

[Sentinel @XriskFYI]

This week, we forecasted about which AI lab will have the top model at the end of the year & whether we are likely to see another pause in inference soon. We looked at the risks of a Russia-NATO military incident, a US ground operation in Iran, & a large plague outbreak. 2/2

[Sentinel @XriskFYI]

The October 5 Sentinel report is out!
Global Risks Weekly Roundup #40/2026
Open model reaches Mythos hacking level, Middle East crude oil flows nearly at pre-Iran-war levels, worker at a Russian plague lab in Siberia died
1/2
https://t.co/ac2omT5jzj