2026-11-06 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 ArchiveModelsAboutCorrectionsThe numbersRSS
My notis
TODAY'S NEWS
November 2026 3 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

Days with a dot have news. See the archive.

The whole week ›
10:40
0/3
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Models being retired Models being withdrawn, and gaps before a replacement is ready.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
Safety 6 NOV

OpenAI rates GPT-6 Cyber Critical for cybersecurity and limits access to approved customers

CodexCybersecurityDaybreak RedOpenAIPreparedness Framework
Illustration: A checkpoint gate with a barrier arm and a reader panel; the green arm and panel stand for access limited to approved customers.Illustration made with AI for ai notis

OpenAI says GPT-6 Cyber, a variant of GPT-6 Astra for authorized security work, produced evidence of 81 high-severity Firefox flaws; only approved Daybreak Red customers have access.

SOURCEOpenAI · 3 sources
CHECKED6 Nov, 05:34
TOPICSFrontier models behind gated programmes, Frontier risk disclosures, Task-specific variants of a frontier family
Safety read
Illustration: A checkpoint gate with a barrier arm and a reader panel; the green arm and panel stand for access limited to approved customers. Safety / 6 NOV

OpenAI rates GPT-6 Cyber Critical for cybersecurity and limits access to approved customers
  • OpenAI classifies GPT-6 Cyber as Critical in cybersecurity under its Preparedness Framework, up from High for GPT-5.6 Cyber, and announced the model on 5 November 2026.
  • OpenAI says GPT-6 Cyber scores 48.5% on ExploitGym, up from 32.1% for GPT-5.6 Cyber, and avoided exploiting out-of-scope services in 99.93% of cases without deployment safety prompts.
  • Access is through the API and Codex for approved Daybreak Red customers, with IP allowlisting required by default; OpenAI says it will widen access to all Daybreak Red customers soon.

OpenAI says its new GPT-6 Cyber model is Critical in cybersecurity under its Preparedness Framework, up from High for GPT-5.6 Cyber, and is giving it first to a select group of approved Daybreak Red customers. OpenAI announced the model on 5 November 2026 in a product post and a system card.

On ExploitGym, which tests whether a model can exploit a specified vulnerability, OpenAI says GPT-6 Cyber scores 48.5%, against 32.1% for GPT-5.6 Cyber. In a separate test where easy-to-attack side services sit next to the task, OpenAI says it avoided exploiting out-of-scope services in 99.93% of cases without its deployment prompt, up from 94.49% for GPT-5.6 Cyber.

OpenAI says its researchers used the model to find 81 previously unknown Firefox vulnerabilities that Mozilla classed as high severity, and to develop working exploits for 948 Linux kernel vulnerabilities. Mozilla's own advisory for Firefox 159, announced on 27 October, credits an OpenAI researcher on several high-impact entries. OpenAI says it disclosed 304 critical Linux findings to maintainers with candidate patches.

Access is through the API and Codex for approved Daybreak Red customers, with IP allowlisting required by default. OpenAI says it will widen access to all Daybreak Red customers soon.

What it means for you Our view

Critical is OpenAI's own rating under its own framework, and the benchmark scores and the Firefox and Linux findings are OpenAI's accounts of its own evaluations. Mozilla's advisory for Firefox 159 adds that an OpenAI researcher is credited on several high-impact entries. For most readers nothing changes today, because only approved Daybreak Red customers can use the model. If your security team uses Daybreak Red or plans to, review whether your networks can meet the IP allowlisting requirement before OpenAI widens access.

TAGSCodexCybersecurityDaybreak RedOpenAIPreparedness Framework
OpenAI · 3 sources tap to close
Tooling 6 NOV

Claude can now see linked bank accounts, read-only, in a US beta

AnthropicClaudePersonal FinancePlaid
A phone screen showing the Personal Finance beta in Claude, with prompts asking what is left after bills and what is in food and drink spending, above a categories chart.Image: Anthropic

Anthropic's Personal Finance beta links bank, card, loan and investment accounts through Plaid for US Pro, Max and Ultra users; Anthropic says hub chats are not used to train its models.

SOURCEAnthropic · 2 sources
CHECKED6 Nov, 05:34
TOPICSWhether consumer and API data train the vendor's models by default
Tooling read
A phone screen showing the Personal Finance beta in Claude, with prompts asking what is left after bills and what is in food and drink spending, above a categories chart. Tooling / 6 NOV

Claude can now see linked bank accounts, read-only, in a US beta
  • Personal Finance is a beta in Claude on the Pro, Max and Ultra plans, in the US only, according to Anthropic's release notes dated 5 November 2026.
  • Anthropic says access is read-only: Personal Finance cannot move, manage or invest money, and is not designed to give investment advice.
  • For the accounts a user connects, Anthropic says Claude sees bank and account names, account type and last four digits, balances, credit limits and up to two years of transactions.

Claude users on Pro, Max and Ultra plans in the US can now link their bank, credit card, loan and investment accounts and ask Claude questions about their own money. Anthropic announced the Personal Finance beta on 5 November 2026 in a blog post and its release notes.

Access is read-only: Anthropic says Personal Finance cannot move, manage or invest money, and is not designed to give investment advice. Users sign in to their bank through Plaid, so Claude does not see the bank password.

For the accounts a user connects, Anthropic says Claude sees the bank and account names, account type and last four digits, balances, credit limits and up to two years of transactions. Conversations in the Personal Finance hub are not used to train its models.

Anthropic's example is a user asking where their money went in a month. It says Claude can total spending by category, flag charges that hit twice, and list payments due before payday.

What it means for you Our view

This is a consumer feature, and what it asks of you is access to your own financial history. Anthropic describes a read-only view, with sign-in to the bank through Plaid so that Claude does not see the bank password, and says conversations in the Personal Finance hub are not used for training. The claims do not cover chats outside the hub. Claude sees only the accounts you choose to connect, so the choice of accounts is yours to make. Before relying on it, connect one account, ask a question you can check against your own statement, and compare the answer.

TAGSAnthropicClaudePersonal FinancePlaid
Anthropic · 2 sources tap to close
toned down Work 6 NOV

Airbnb reports about 40% of support issues were resolved without a human agent

AI assistantAirbnbCustomer support
Illustration: A sorting station with an intake slot at the top, a green chute running to an outbox tray, and a separate round gauge dial on its side panel.Illustration made with AI for ai notis

Airbnb's Q3 letter says support cost per booking fell more than 17%, driven in large part by improvements to its AI assistant, as issues resolved without a human rose from a third in Q2.

SOURCEAirbnb, Inc. · 1 source
CHECKED6 Nov, 05:34
Work read
Illustration: A sorting station with an intake slot at the top, a green chute running to an outbox tray, and a separate round gauge dial on its side panel. Work / 6 NOV

Airbnb reports about 40% of support issues were resolved without a human agent
  • Airbnb says approximately 40% of all customer support issues in Q3 2026, not just those that begin in its AI assistant, were resolved without a human agent.
  • The letter says support cost per booking fell more than 17% year over year, driven in large part by improvements to the AI assistant, now in more than 50 languages.
  • Airbnb says it is building its own AI agent to plan and book trips and plans to start rolling it out to guests in the first half of 2027.

About 40% of all customer support issues at Airbnb were resolved without a human agent in the third quarter of 2026, up from about a third in the second quarter, according to the company's Q3 2026 shareholder letter.

The letter, filed with the SEC as an exhibit to a Form 8-K, says customer support cost per booking fell more than 17% year over year. It also says Airbnb's AI assistant is available in more than 50 languages.

Airbnb also says it is building its own AI agent to help guests plan and book trips, and plans to start rolling it out to guests in the first half of 2027. That is a plan, not a launch.

The letter reports third-quarter revenue of $4.9 billion, up 19% year over year.

What it means for you Our view

The letter does not say the AI assistant resolved those issues: it counts all support issues, whatever began in the assistant. The cost fall is linked to the assistant in Airbnb's own words, so read it as the company's attribution. The trip-planning agent is a plan for 2027 and has not launched. If you are weighing a support assistant, ask which issues it settles end to end and which only reach a person later, and watch Airbnb's next quarterly letter to see whether the share without a human keeps rising.

TAGSAI assistantAirbnbCustomer support
Airbnb, Inc. · 1 source tap to close
NOTHING IN THIS FILTER No story today matches. open the archive →
That's today's news: three stories. ExplainedHow a system card reports a model's tests, ratings and access limits

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 3
Illustration: A checkpoint gate with a barrier arm and a reader panel; the green arm and panel stand for access limited to approved customers. Illustration made with AI for ai notis

Safety · 6 NOV

OpenAI rates GPT-6 Cyber Critical for cybersecurity and limits access to approved customers

OpenAI says GPT-6 Cyber, a variant of GPT-6 Astra for authorized security work, produced evidence of 81 high-severity Firefox flaws; only approved Daybreak Red customers have access.

  • OpenAI classifies GPT-6 Cyber as Critical in cybersecurity under its Preparedness Framework, up from High for GPT-5.6 Cyber, and announced the model on 5 November 2026.
  • OpenAI says GPT-6 Cyber scores 48.5% on ExploitGym, up from 32.1% for GPT-5.6 Cyber, and avoided exploiting out-of-scope services in 99.93% of cases without deployment safety prompts.
  • Access is through the API and Codex for approved Daybreak Red customers, with IP allowlisting required by default; OpenAI says it will widen access to all Daybreak Red customers soon.
Checked 6 Nov, 05:34 · OpenAI · 3 sources
Read the story
A phone screen showing the Personal Finance beta in Claude, with prompts asking what is left after bills and what is in food and drink spending, above a categories chart. Image: Anthropic

Tooling · 6 NOV

Claude can now see linked bank accounts, read-only, in a US beta

Anthropic's Personal Finance beta links bank, card, loan and investment accounts through Plaid for US Pro, Max and Ultra users; Anthropic says hub chats are not used to train its models.

  • Personal Finance is a beta in Claude on the Pro, Max and Ultra plans, in the US only, according to Anthropic's release notes dated 5 November 2026.
  • Anthropic says access is read-only: Personal Finance cannot move, manage or invest money, and is not designed to give investment advice.
  • For the accounts a user connects, Anthropic says Claude sees bank and account names, account type and last four digits, balances, credit limits and up to two years of transactions.
Checked 6 Nov, 05:34 · Anthropic · 2 sources
Read the story
Illustration: A sorting station with an intake slot at the top, a green chute running to an outbox tray, and a separate round gauge dial on its side panel. Illustration made with AI for ai notis

Work · 6 NOVnarrowed

Airbnb reports about 40% of support issues were resolved without a human agent

Airbnb's Q3 letter says support cost per booking fell more than 17%, driven in large part by improvements to its AI assistant, as issues resolved without a human rose from a third in Q2.

  • Airbnb says approximately 40% of all customer support issues in Q3 2026, not just those that begin in its AI assistant, were resolved without a human agent.
  • The letter says support cost per booking fell more than 17% year over year, driven in large part by improvements to the AI assistant, now in more than 50 languages.
  • Airbnb says it is building its own AI agent to plan and book trips and plans to start rolling it out to guests in the first half of 2027.
Checked 6 Nov, 05:34 · Airbnb, Inc. · 1 source
Read the story

You're all caught up

That's today's news: three stories.

ExplainedHow a system card reports a model's tests, ratings and access limits
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
Illustration: A checkpoint gate with a barrier arm and a reader panel; the green arm and panel stand for access limited to approved customers. Illustration made with AI for ai notis Safety · 6 NOV OpenAI rates GPT-6 Cyber Critical for cybersecurity and limits access to approved customers OpenAI · 3 sources · 6 Nov, 05:34 A phone screen showing the Personal Finance beta in Claude, with prompts asking what is left after bills and what is in food and drink spending, above a categories chart. Image: Anthropic Tooling · 6 NOV Claude can now see linked bank accounts, read-only, in a US beta Anthropic · 2 sources · 6 Nov, 05:34 Illustration: A sorting station with an intake slot at the top, a green chute running to an outbox tray, and a separate round gauge dial on its side panel. narrowedIllustration made with AI for ai notis Work · 6 NOV Airbnb reports about 40% of support issues were resolved without a human agent Airbnb, Inc. · 1 source · 6 Nov, 05:34

That's today's news: 3 stories.Archive

Today's stories 3 STORIES

Today's stories

Fri 6 Nov · 3 STORIES
  1. Safety · 6 NOV

    OpenAI rates GPT-6 Cyber Critical for cybersecurity and limits access to approved customers

    Illustration: A checkpoint gate with a barrier arm and a reader panel; the green arm and panel stand for access limited to approved customers.Illustration made with AI for ai notis

    OpenAI says GPT-6 Cyber, a variant of GPT-6 Astra for authorized security work, produced evidence of 81 high-severity Firefox flaws; only approved Daybreak Red customers have access.

    Checked 6 Nov, 05:34 · OpenAI · 3 sourcesRead

  2. Tooling · 6 NOV

    Claude can now see linked bank accounts, read-only, in a US beta

    A phone screen showing the Personal Finance beta in Claude, with prompts asking what is left after bills and what is in food and drink spending, above a categories chart.Image: Anthropic

    Anthropic's Personal Finance beta links bank, card, loan and investment accounts through Plaid for US Pro, Max and Ultra users; Anthropic says hub chats are not used to train its models.

    Checked 6 Nov, 05:34 · Anthropic · 2 sourcesRead

  3. Work · 6 NOV · narrowed

    Airbnb reports about 40% of support issues were resolved without a human agent

    Illustration: A sorting station with an intake slot at the top, a green chute running to an outbox tray, and a separate round gauge dial on its side panel.Illustration made with AI for ai notis

    Airbnb's Q3 letter says support cost per booking fell more than 17%, driven in large part by improvements to its AI assistant, as issues resolved without a human rose from a third in Q2.

    Checked 6 Nov, 05:34 · Airbnb, Inc. · 1 sourceRead

Safety 6 NOV

OpenAI rates GPT-6 Cyber Critical for cybersecurity and limits access to approved customers

OpenAI says GPT-6 Cyber, a variant of GPT-6 Astra for authorized security work, produced evidence of 81 high-severity Firefox flaws; only approved Daybreak Red customers have access.

OpenAI · 3 sources·checked 6 Nov, 05:34
CodexCybersecurityDaybreak RedOpenAIPreparedness Framework
Illustration: A checkpoint gate with a barrier arm and a reader panel; the green arm and panel stand for access limited to approved customers.Illustration made with AI for ai notis
Tooling

Claude can now see linked bank accounts, read-only, in a US beta

Anthropic's Personal Finance beta links bank, card, loan and investment accounts through Plaid for US Pro, Max and Ultra users; Anthropic says hub chats are not used to train its models.

6 NOV · Anthropic · 2 sources
toned down Work

Airbnb reports about 40% of support issues were resolved without a human agent

Airbnb's Q3 letter says support cost per booking fell more than 17%, driven in large part by improvements to its AI assistant, as issues resolved without a human rose from a third in Q2.

6 NOV · Airbnb, Inc. · 1 source

Explained

One idea behind the news
Safety How a system card reports a model's tests, ratings and access limits OpenAI's deployment safety hub says it shares how deployed models perform in evaluations, the risks measured and the steps taken; the GPT-6 Cyber card of 5 November 2026 shows the layout.
In brief 2 more today
0 of 3 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