2026-10-01 notis
OTHER NOTIS SITES AIYOU ARE HERE AI models, prices and agents. ainotis.com Hiking Trail openings, hut bookings, and route conditions. hiking.notis.news Photography Cameras, lenses and firmware. photo.notis.news More areas when we can cover them every day.
CATEGORY Join
Show Read as Find Sections Models and releasesAgents and appsResearchMoney and computeLaws and rulingsSafety and securityAI at work More The monthlyArchiveModelsAboutCorrectionsThe numbersRSS
My notis
TODAY'S NEWS
October 2026 2 STORIES
MTWTFSS
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31

Days with a dot have news. See the archive.

The whole week ›
15:45
0/2
WEATHER THEME CARD SPACING

Today's news

today

My page

Kept in this browser only.

SAVED FOLLOWING COLLECTIONS READING SINCE
SAVED STORIES
Nothing saved yet. Tap the bookmark on a card to keep it here. read today
CASES YOU FOLLOW all cases

No case followed yet.

COLLECTIONS
Price changes Stories where a published price went up or down.
Models being retired Models being withdrawn, and gaps before a replacement is ready.
THE MORNING EMAIL The day's top stories. Sign up
toned down Models 1 OCT

Google releases Gemini 4 Argon first to trusted cyber defenders

GeminiGooglegemini-4-argon
Release artwork on a blue gradient: the Gemini star and the name Gemini 4 Argon on the left, a large white numeral 4 on the right.Image: Google

Google announced Gemini 4 Argon on 30 September, rolling it out first to trusted cyber defenders through its Fairwind Program, with wider access to follow.

SOURCEGoogle · 1 source
CHECKED1 Oct, 04:18
TOPICSFrontier models behind gated programmes, Maximum output limits rising
Models read
Release artwork on a blue gradient: the Gemini star and the name Gemini 4 Argon on the left, a large white numeral 4 on the right. Models / 1 OCT

Google releases Gemini 4 Argon first to trusted cyber defenders
  • Google says Argon is rolling out to a set of trusted cyber defenders through its Fairwind Program, and will be released to them without cyber guardrails.
  • Google says Argon launches at an introductory price of $2 per million input tokens and $10 per million output tokens, and its output limit rises to 1M tokens from 64K.
  • Google reports 77.9% on DeepSWE v1.1 and 91.7% on LVBench, calling both state of the art, and says Argon ranks first on Zapier’s AutomationBench at 51.3%.

Google announced Gemini 4 Argon on 30 September 2026. It is rolling out first to a set of trusted cyber defenders through the Fairwind Program. Google says it will make Argon available to developers, enterprises and consumers as soon as possible, after a phased expansion. Argon will launch at an introductory price of $2 per million input tokens and $10 per million output tokens. Cached input tokens are priced at 95% off the input token price. Google raised the output token limit to 1M tokens, up from 64K. Google reports 77.9% on DeepSWE v1.1, 51.3% on AutomationBench and 91.7% on LVBench. For trusted defenders and its own teams, Google will release Argon without cyber guardrails.

What it means for you Our view

Google says Argon reaches developers, enterprises and consumers after the trusted defenders, starting with paid API customers and Google AI Ultra subscribers, and that early feedback feeds its work on guardrails. The prices it gives are introductory, so a budget built on them needs a second figure from Google for the period afterwards. The scores and the label state of the art are Google’s own, and the AutomationBench ranking is Google’s report on a third party’s test. If you plan to use Argon, ask Google what price applies after the introductory period, and run it on your own tasks before relying on its published results.

TAGSGeminiGooglegemini-4-argon
Google · 1 source tap to close
toned down Research 1 OCT

Anthropic says GLM-5.3 builds exploits close to Claude Mythos Preview

AnthropicZ.aiclaude-mythos-previewglm-5-3
Two line charts of exploitation success against output-token budget: on ExploitBench Claude Mythos Preview reaches 14 percent and GLM-5.3 12 percent, other models near 0.Image: Anthropic

Anthropic’s Frontier Red Team published a 29 September analysis of Zhipu AI’s open-weight GLM-5.3, covering its own tests of the model’s cyber capability and safeguards.

SOURCEAnthropic · 1 source
CHECKED1 Oct, 04:18
TOPICSGovernment model evaluations, Open weights at frontier scale, Safeguards for open-weight release
Research read
Two line charts of exploitation success against output-token budget: on ExploitBench Claude Mythos Preview reaches 14 percent and GLM-5.3 12 percent, other models near 0. Research / 1 OCT

Anthropic says GLM-5.3 builds exploits close to Claude Mythos Preview
  • On ExploitBench, Anthropic reports GLM-5.3 developed end-to-end exploits in 50 of 410 attempts and Claude Mythos Preview in 56 of 410, in Anthropic’s own test.
  • Anthropic says in its simulated tests GLM-5.3 engaged with overtly malicious requests to attack critical systems 64% with a deceptive prompt, 92% with prefilled thinking and 100% when abliterated.
  • Anthropic says NIST’s CAISI found GLM-5.3 "the most cyber-capable open-weight model released to date" in an assessment published on 17 September.

Anthropic’s Frontier Red Team published an analysis of Zhipu AI’s GLM-5.3 on 29 September 2026. On ExploitBench, GLM-5.3 developed end-to-end exploits in 50 of 410 attempts, and Claude Mythos Preview did so in 56 of 410. On Anthropic’s internal Binary Exploitation benchmark, GLM-5.3 achieved full control-flow hijacks in 4% of trials and Claude Mythos Preview in 6%. Anthropic says attackers can bypass GLM-5.3’s safeguards between 64% and 100% of the time with simple techniques in its simulated tests, and that these did not succeed against safeguarded Claude models. A researcher using the smaller GLM-5.3-Flash built an exploit chain for a Chrome flaw (CVE-2026-11645) at an API cost Anthropic puts at $20.40. Anthropic also says NIST’s CAISI published its own assessment of GLM-5.3 on 17 September and found it the most cyber-capable open-weight model released to date.

What it means for you Our view

Every figure here comes from Anthropic’s tests of a competitor’s model, so read it as one lab’s view. Within that, Anthropic reports GLM-5.3 close to Claude Mythos Preview on its exploit benchmarks, and says simple techniques bypassed GLM-5.3’s safeguards in its simulated environment, where the bypass rates were measured. The CAISI finding reaches us through Anthropic’s post, not the report itself. If you evaluate open-weight models for security work, ask which safeguards a model ships with, and read the CAISI assessment directly before relying on this comparison.

TAGSAnthropicZ.aiclaude-mythos-previewglm-5-3
Anthropic · 1 source tap to close
NOTHING IN THIS FILTER No story today matches. open the archive →
That's today's news: two stories. This edition carries two leads rather than three; the story we would have placed third is among the quick hits.

The morning email

On the mornings we publish: the three top stories and up to four short ones. Free.

We email you a link to confirm. An issue may include one sponsor, always labelled Sponsored · Advertisement. Our emails count opens and clicks, not who made them. Unsubscribe in one click. What we keep

1 of 2
Release artwork on a blue gradient: the Gemini star and the name Gemini 4 Argon on the left, a large white numeral 4 on the right.Image: Google

Models · 1 OCTnarrowed

Google releases Gemini 4 Argon first to trusted cyber defenders

Google announced Gemini 4 Argon on 30 September, rolling it out first to trusted cyber defenders through its Fairwind Program, with wider access to follow.

  • Google says Argon is rolling out to a set of trusted cyber defenders through its Fairwind Program, and will be released to them without cyber guardrails.
  • Google says Argon launches at an introductory price of $2 per million input tokens and $10 per million output tokens, and its output limit rises to 1M tokens from 64K.
  • Google reports 77.9% on DeepSWE v1.1 and 91.7% on LVBench, calling both state of the art, and says Argon ranks first on Zapier’s AutomationBench at 51.3%.
Checked 1 Oct, 04:18 · Google · 1 source
Read the story
Two line charts of exploitation success against output-token budget: on ExploitBench Claude Mythos Preview reaches 14 percent and GLM-5.3 12 percent, other models near 0.Image: Anthropic

Research · 1 OCTnarrowed

Anthropic says GLM-5.3 builds exploits close to Claude Mythos Preview

Anthropic’s Frontier Red Team published a 29 September analysis of Zhipu AI’s open-weight GLM-5.3, covering its own tests of the model’s cyber capability and safeguards.

  • On ExploitBench, Anthropic reports GLM-5.3 developed end-to-end exploits in 50 of 410 attempts and Claude Mythos Preview in 56 of 410, in Anthropic’s own test.
  • Anthropic says in its simulated tests GLM-5.3 engaged with overtly malicious requests to attack critical systems 64% with a deceptive prompt, 92% with prefilled thinking and 100% when abliterated.
  • Anthropic says NIST’s CAISI found GLM-5.3 "the most cyber-capable open-weight model released to date" in an assessment published on 17 September.
Checked 1 Oct, 04:18 · Anthropic · 1 source
Read the story

You're all caught up

That's today's news: two stories.

Start again See the whole week

You can join ai notis soon, to follow cases and keep stories on every device.

Follow us on Instagram and Threads

The morning email

On the mornings we publish: the three top stories and up to four short ones. Free.

We email you a link to confirm. An issue may include one sponsor, always labelled Sponsored · Advertisement. Our emails count opens and clicks, not who made them. Unsubscribe in one click. What we keep

No story today matches. Open the archive
Release artwork on a blue gradient: the Gemini star and the name Gemini 4 Argon on the left, a large white numeral 4 on the right. narrowedImage: Google Models · 1 OCT Google releases Gemini 4 Argon first to trusted cyber defenders Google · 1 source · 1 Oct, 04:18 Two line charts of exploitation success against output-token budget: on ExploitBench Claude Mythos Preview reaches 14 percent and GLM-5.3 12 percent, other models near 0. narrowedImage: Anthropic Research · 1 OCT Anthropic says GLM-5.3 builds exploits close to Claude Mythos Preview Anthropic · 1 source · 1 Oct, 04:18

That's today's news: 2 stories.Archive

Today's stories 2 STORIES

Today's stories

Thu 1 Oct · 2 STORIES
  1. Models · 1 OCT · narrowed

    Google releases Gemini 4 Argon first to trusted cyber defenders

    Release artwork on a blue gradient: the Gemini star and the name Gemini 4 Argon on the left, a large white numeral 4 on the right.Image: Google

    Google announced Gemini 4 Argon on 30 September, rolling it out first to trusted cyber defenders through its Fairwind Program, with wider access to follow.

    Checked 1 Oct, 04:18 · Google · 1 sourceRead

  2. Research · 1 OCT · narrowed

    Anthropic says GLM-5.3 builds exploits close to Claude Mythos Preview

    Two line charts of exploitation success against output-token budget: on ExploitBench Claude Mythos Preview reaches 14 percent and GLM-5.3 12 percent, other models near 0.Image: Anthropic

    Anthropic’s Frontier Red Team published a 29 September analysis of Zhipu AI’s open-weight GLM-5.3, covering its own tests of the model’s cyber capability and safeguards.

    Checked 1 Oct, 04:18 · Anthropic · 1 sourceRead

toned down Models 1 OCT

Google releases Gemini 4 Argon first to trusted cyber defenders

Google announced Gemini 4 Argon on 30 September, rolling it out first to trusted cyber defenders through its Fairwind Program, with wider access to follow.

Google · 1 source·checked 1 Oct, 04:18
GeminiGooglegemini-4-argon
Release artwork on a blue gradient: the Gemini star and the name Gemini 4 Argon on the left, a large white numeral 4 on the right.Image: Google
toned down Research

Anthropic says GLM-5.3 builds exploits close to Claude Mythos Preview

Anthropic’s Frontier Red Team published a 29 September analysis of Zhipu AI’s open-weight GLM-5.3, covering its own tests of the model’s cyber capability and safeguards.

1 OCT · Anthropic · 1 source
In brief 3 more today

This edition carries two leads rather than three; the story we would have placed third is among the quick hits.

0 of 2 read
Drag a card to start Drag left or right · tap to read more Hover to lift a card · tap to bring it forward Tap again to put it back Tap a card to open it Tap again to close it Tap a headline to read it as a card