Scaling Trusted Access for Cyber with GPT‑5.5 and GPT‑5.5‑Cyber

For years we’ve been chronicling our work to accelerate cybersecurity defenders, as part of our broader work to build the core infrastructure for AI. Last week, we released our action plan Cybersecurity in the Intelligence Age, which lays out our vision for democratizing AI-powered defense. Two weeks ago, we released GPT‑5.5, our smartest and most intuitive model to date, which is already delivering powerful cybersecurity capabilities to developers and security teams through Trusted Access for Cyber (TAC).

Today, we are rolling out GPT‑5.5‑Cyber in limited preview to defenders responsible for securing critical infrastructure to support specialized cybersecurity workflows that help protect the broader ecosystem.

We are focused on providing proportional safeguards and access to empower cyber defenders to protect society, and our approach has been informed by conversations with cybersecurity and national security leaders across federal and state government and major commercial entities.

The cyber defense ecosystem is broad, and GPT‑5.5 and GPT‑5.5‑Cyber play different roles in meeting the needs of organizations and researchers across it, depending on the task, the setting, and the safeguards around how the model is used. For most teams, GPT‑5.5 with TAC is our strongest broadly useful model for legitimate defensive work, with strong safeguards against misuse.

In this post, we are sharing more details on how Trusted Access for Cyber works, how GPT‑5.5 and GPT‑5.5‑Cyber meet the varied needs of defenders across the ecosystem, and how different levels of access affect model outputs.

Trusted Access for Cyber is an identity and trust-based framework designed to help ensure enhanced cyber capabilities are being placed in the right hands. It is designed to make the cyber capabilities of GPT‑5.5 more useful for verified defenders working on defensive tasks, while continuing to restrict requests that could enable real-world harm.

When defenders are vetted and approved for Trusted Access for Cyber, they receive lower classifier-based refusals to enable authorized cybersecurity workflows, including vulnerability identification and triage, malware analysis, binary reverse engineering, detection engineering, and patch validation. Safeguards continue to block malicious activity such as credential theft, stealth, persistence, malware deployment, or exploitation of third-party systems.

As we announced last week, with increased access, defenders are required to have phishing-resistant account security protections. Individual members of Trusted Access for Cyber accessing our most cyber capable and permissive models will be required to enable Advanced Account Security beginning June 1, 2026. Organizations with trusted access can, as an alternative, attest that they have phishing resistant authentication as part of their single sign-on workflow.

Here is a breakdown for how to think about the current trusted access levels:

The differences between model access levels are most pronounced when comparing prompts and responses. The first example illustrates how GPT‑5.5 compares to GPT‑5.5 with Trusted Access for Cyber on a defensive task: create a proof-of-concept from a published vulnerability to validate remediation within an authorized environment.

For most defenders, GPT‑5.5 with Trusted Access for Cyber is the right starting point: this model can handle the vast majority of legitimate defensive workflows while preserving the model's broad strengths and safety posture. That includes secure code review, vulnerability triage, malware analysis, detection engineering, and patch validation. 

More specialized access becomes relevant only when authorized workflows still run into refusals. This occurs with higher risk workflows such as red teaming and penetration testing, where defenders may need to go beyond analysis, and validate exploitability in a controlled environment. GPT‑5.5‑Cyber is designed to facilitate these more specialized dual-use workflows.

Here’s a simple example that shows what that looks like in practice:

GPT‑5.5 ndiyo muundo wetu wenye akili zaidi na unaoeleweka kwa urahisi zaidi kwa kazi za maarifa za matumizi ya jumla na kazi za usalama wa mtandao, na ndiyo muundo tunaotarajia walinzi wengi watumie. Tunatathmini utendaji wa mtandao katika kazi zinazohitaji uwazaji wa hatua nyingi, matumizi ya zana, na ustahimilivu katika mtiririko halisi wa kazi za ulinzi.

Onyesho la kwanza la miundo inyoruhusu zaidi kazi za mtandao kama GPT‑5.5‑Cyber halikusudiwi kuongeza kwa kiasi kikubwa uwezo wa mtandao zaidi ya GPT‑5.5 - kimsingi limefundishwa kuwa na ruhusa zaidi kwenye kazi zinazohusiana na usalama.

Kwa hiyo, onyesho hili la kwanza halitarajiwi kuizidi GPT‑5.5 katika kila tathmini ya mtandao. Badala yake, linaunga mkono mchakato wa utekelezaji wa hatua kwa hatua ili kuharakisha walinzi na kusaidia kwa usalama mtiririko maalum zaidi wa kazi zilizoidhinishwa unaohitaji tabia yenye ruhusa zaidi, ukiambatana na uthibitishaji imara zaidi, ufuatiliaji wa matumizi mabaya, uwekaji mipaka wa matumizi yaliyoidhinishwa, na mrejesho wa washirika. Kwa sasa, GPT‑5.5 yenye Trusted Access for Cyber bado ndiyo sehemu ya kuanzia inayopendekezwa kwa mtiririko mingi ya kazi za usalama.

Tunashirikiana na wauzaji wa usalama kwa sababu wako mahali ambapo uwezo wa muundo unaweza kuwa ulinzi wa mteja: ugunduzi, uundaji, utambuzi, mwitikio, na utekelezaji wa mtandao. Tabaka hizo zinapoboreshwa pamoja, huunda mzunguko wa usalama: watafiti hufichua dosari pamoja na proof-of-concept za matumizi mabaya na mwongozo wa viraka, zana za mnyororo wa ugavi wa programu huzuia msimbo wenye dosari na dependencies zilizoathiriwa zisifike uzalishoni, washirika wa EDR na SIEM hugundua utumiaji porini, na watoa huduma wa mtandao na usalama hupeleka hatua za kupunguza katika kiwango cha WAF huku marekebisho yakiendelea kutolewa.

GPT‑5.5 yenye Trusted Access for Cyber ndiyo msingi mpana wa kazi hii. Inaweza kusaidia walinzi waliothibitishwa kusonga kwa kasi zaidi katika mzunguko wa maisha wa usalama, huku GPT‑5.5‑Cyber ikiruhusu kundi dogo la washirika kuchunguza mtiririko wa hali ya juu ambapo tabia maalum ya ufikiaji inaweza kuwa muhimu. Lengo ni kusaidia mfumo wa usalama kulinda wateja kwa haraka zaidi, kisha kujifunza kutoka kwa mrejesho wa washirika mahali ambapo tathmini, uthibitishaji, au kinga kali zaidi zinahitajika.

Watoa huduma wa mtandao na usalama

Watoa huduma wa mtandao na usalama wanaweza kupunguza exposure huku marekebisho yakiwa bado yanatolewa. Walinzi wanapothibitisha dosari na kufuatilia iwapo inatumika, wanaweza pia kuweka sheria za WAF, hatua za ukingoni za kupunguza, na mabadiliko ya usanidi yanayodhoofisha njia za mashambulizi zinazowezekana kabla kila mfumo ulioathiriwa haujarekebishwa. GPT‑5.5 inaweza kusaidia ukaguzi wa sheria, uchambuzi wa usanidi, uchunguzi wa matukio, na usimamizi salama wa mabadiliko katika mazingira changamano. 

Tunafanya kazi na washirika hawa ili kutusaidia kutathmini jinsi uwezo huo unavyotafsiriwa kuwa ulinzi ambao wateja wanaweza kutekeleza kwa kiwango cha intaneti, ikiwemo kwa miundomsingi muhimu na huduma za umma ambapo kupunguza exposure haraka ni muhimu.

Utafiti wa dosari na kuweka viraka

Mzunguko huanza kwa kupata dosari, kuthibitisha umuhimu wake, na kuweka viraka kwenye mifumo iliyoathiriwa. GPT‑5.5 yenye Trusted Access for Cyber inaweza kusaidia sehemu kubwa ya kazi hii: kuelewa msimbo usiozoeleka, kupanga ramani ya maeneo yaliyoathiriwa, kufuatilia chanzo kikuu, kukagua viraka, kujenga zana salama za kuiga hali, kuweka kipaumbele cha ukali, na kubadilisha matokeo kuwa mwongozo wa urekebishaji.

Baadhi ya utafiti wa dosari unahitaji tabia yenye ruhusa zaidi, hasa pale washirika walioidhinishwa wanapohitaji proof-of-concept za matumizi mabaya kwa ajili ya ufichuzi ulioratibiwa au uthibitishaji uliofanyiwa udhibiti. Huo ndio mtiririko wa kazi ambapo GPT‑5.5‑Cyber inaweza kutusaidia kujifunza na kundi dogo la washirika, chini ya uthibitishaji, ufuatiliaji, na mizunguko ya mrejesho iliyo imara zaidi.

Utambuzi na ufuatiliaji

Iwapo programu yenye dosari tayari imetekelezwa, swali linalofuata ni kama kuna yeyote anayeitumia vibaya. Washirika wa EDR, SIEM, IGA/PAM, na ufuatiliaji hubadilisha tahadhari mpya kuwa ushahidi kutoka mazingira ya moja kwa moja: telemetry, arifa, utambuzi, na mtiririko wa mwitikio. GPT‑5.5 inaweza kusaidia wachambuzi kuunganisha ishara hizo, kufupisha yaliyo muhimu, kuandaa utambuzi, na kusonga kwa haraka zaidi kutoka ufichuzi hadi uchunguzi. Mzunguko huo huo ni muhimu hasa katika mazingira ya wingu, ambako exposure, urekebishaji, na utambuzi vimeunganishwa kwa karibu.

Usalama wa mnyororo wa ugavi wa programu

Mzunguko unaofuata ni kuzuia msimbo unaojulikana kuwa mbaya usifike uzalishoni tangu mwanzo. Mara dosari au uharibifu wa kifurushi unapoeleweka, zana za mnyororo wa ugavi wa programu zinaweza kusaidia kuzuia dependencies hatarishi, masasisho hasidi, na njia za msimbo zenye dosari kabla hazijasambaa katika mazingira ya wateja. GPT‑5.5 yenye Trusted Access for Cyber inaweza kusaidia kukagua mabadiliko ya dependency, kufanya uwazaji kuhusu uwezekano wa matumizi mabaya katika msimbo unaomilikiwa, kuweka kipaumbele cha urekebishaji, na kuibua mapema tabia za kifurushi zinazotiliwa shaka katika mzunguko wa maendeleo. 

Washirika kama Snyk, Gen Digital, Semgrep, na Socket wanaweza kutusaidia kujaribu jinsi uwezo huu unavyotumika katika matukio kama uharibifu wa axios, ambapo suluhisho la haraka zaidi ni kuzuia dependencies zenye dosari au zilizoathiriwa zisiingie kabisa katika build.

Open source ni mojawapo ya njia za haraka zaidi ambazo dosari inaweza kusambaa katika mfumo mzima, hivyo pia tunawekeza juu ya mkondo pamoja na wahudumu wa matengenezo. Codex Security husaidia timu kutambua, kuthibitisha, na kurekebisha dosari kwa kujenga muundo wa vitisho mahsusi kwa codebase, kuchunguza njia halisi za mashambulizi, kuthibitisha masuala katika mazingira yaliyotengwa, na kupendekeza viraka kwa ukaguzi wa binadamu.

Kupitia Codex for Open Source, wahudumu walioteuliwa wa miradi muhimu wanaweza kupata ufikiaji wa masharti kwa Codex Security pamoja na Codex na mikopo ya API ili kupunguza mzigo wa matengenezo na ukaguzi.

Pia tumetoa programu-jalizi ya Codex Security⁠ inayoleta mtiririko uliopo wa kazi za usalama moja kwa moja kwenye kiolesura chochote cha Codex kama programu au CLI, ikiwasaidia wasanidi kusonga kutoka uundaji wa muundo wa vitisho hadi ugunduzi, uthibitishaji, uchambuzi wa njia za mashambulizi, na marekebisho yaliyothibitishwa.

Miundo inapokuwa na uwezo zaidi katika usalama wa mtandao, matumizi bora ya uwezo huo ni kuwasaidia walinzi kupata na kurekebisha udhaifu kwa haraka zaidi. Kupanua ufikiaji kwa uwezo huo kwa uwajibikaji kunahitaji uaminifu mkubwa zaidi kuhusu nani anayetumia muundo, ni mifumo gani wanayolenga, na kama kazi hiyo imeidhinishwa. Kadiri uthibitishaji imara zaidi wa utambulisho na wa mashirika, uwekaji mipaka wa matumizi yaliyoidhinishwa, na ufuatiliaji wa matumizi mabaya unavyoboreshwa, tunatarajia ufikiaji kupanuka baada ya muda.

Kupata ufikiaji wa Trusted Access for Cyber ni rahisi:

• Watumiaji binafsi wanaweza kuthibitisha utambulisho wao katika chatgpt.com/cyber⁠⁠(fungua katika dirisha jipya).
• Biashara zinaweza kuomba ufikiaji wa kuaminika⁠ kwa timu yao kupitia mwakilishi wao wa OpenAI.

Wateja wote watakaoidhinishwa kupitia mchakato huu watapata ufikiaji wa matoleo ya miundo iliyopo yenye msuguano mdogo zaidi kuhusu kinga ambazo zinaweza kuwashwa na shughuli za mtandao zenye matumizi mawili, na kuwaruhusu kuendelea kuunga mkono elimu ya usalama, upangaji wa ulinzi, na utafiti wa dosari kwa uwajibikaji. 

Wakati wa majaribio ya alpha, GPT‑5.5‑Cyber tayari imetumika kupanua red teaming ya kiotomatiki ya mifumo muhimu na kuthibitisha dosari zenye ukali wa juu, jambo ambalo tutaliandika katika uchambuzi wa kina wa kiufundi wa siku zijazo kama sehemu ya ufichuzi wa uwajibikaji.

Tunatarajia kuendelea kuharakisha walinzi kwa miundo mbalimbali, ikiwemo miundo yetu mkuu kupitia Trusted Access for Cyber, pamoja na miundo maalum ya mtandao kama GPT‑5.5‑Cyber na hata miundo yenye uwezo mkubwa zaidi wa mtandao katika siku zijazo.