Anthropic Waxay tidhi Fable 5 Dib-u-dhacyada Biology-ga ayaa hoos u dhacay 85%

Maxaa dhacay

Anthropic waa la cusboonaysiiyay Claude Sheekada 5 ilaalinta bayoolaji ee Agoosto 7, iyadoo la soo koobay kala soocida ku dhawaad su'aal kasta oo bayoolooji ah u rogtay nooc ka awood yar bayoloji.

Marka kalasaaruhu calaamadiyo codsi, Anthropic wuxuu ka soo qaadaa Fable 5 ilaa Opus 5. Shirkaddu waxay sheegtay in Opus 5 uu weli awood u leeyahay isticmaalka guud laakiin wuxuu bixiyaa caawimo hawleed yar oo ku saabsan bayoolaji horumarsan, hoos u dhigista qiimaha nidaamka qof raacaya shaqo waxyeello leh.

Anthropic waxay sheegtay inay dib u qortay dastuurkii kala soocida, waxay soo ururisay jawaab celin khubarada gudaha iyo dibadaba, waxay abuurtay xog tababar cusub, dib u tayaysay kalasaarkii, waxayna hubisay inay wali guud ahaan kacday codsiyada cilmi baadhiseed ee waxyeelada leh iyo labada isticmaalba. Tijaabada shirkadda, cusboonaysiinta ayaa hoos u dhigtay dib-u-dhacyada la xidhiidha bayooloji ilaa 85% dhammaan alaabteeda.

The change follows a deliberately conservative launch posture. Anthropic says Fable 5 initially routed almost every biology request to Opus 5 because the company preferred a broad safety boundary while it learned where benign health and education questions were being caught. The August update narrows that boundary through a separate classifier instead of changing the model's underlying biology capability. That makes the release a policy-and-routing change, not evidence that Fable 5 has become a clinically validated biology assistant.

Isbeddelka waxaa loola jeedaa in Fable 5 uu ka jawaabo su'aalo badan oo caafimaad, caafimaad, iyo waxbarasho maalinle ah. Anthropic waxay sheegtay in gelitaanka caadiga ah ay wali dib ugu dhacayso aagagga la isticmaalo ee ay ka mid yihiin virology, toxicology, iyo design molecular, marka moodeelka wali lama heli karo iyada oo loo marayo dariiqa cilmi baarista bayoolajiga xirfadlaha ah ama horumarinta daroogada.

Faahfaahinta isha: Anthropic's Fable 5 biology safeguards announcement ↗

Maxay muhiim u tahay

Cusboonaysiinta ayaa ah tijaabo wax ku ool ah oo ku saabsan in ilaalinta moodooyinka soohdintu ay noqon karaan kuwo sax ah iyada oo aan si fudud loo kala dooran marin ballaadhan iyo diidmo ballaadhan.

Shaandheeyaha aan dhab ahayn wuxuu yarayn karaa khatarta si dhakhso ah, laakiin sidoo kale waxay xannibi kartaa ardayda, bukaanka, barayaasha, iyo xirfadlayaasha daryeelka caafimaadka kuwaas oo su'aalahooda ay isticmaalaan luqad farsamo oo isku mid ah sida cilmi-baaris xasaasi ah. Anthropic waxay dooratay bartaas muxaafidka ah markii ay sii daysay Fable 5, ka dib waxay adeegsatay siyaasad aad u faahfaahsan iyo tusaalooyin tababar oo cusub si loogu wareejiyo xuduudka codsiyada aan fiicnayn.

Isbeddelka waayo-aragnimada isticmaaluhu wuu ku kala duwan yahay badeecada sababtoo ah bayoolaji waa hal sabab oo ah dib u dhac. Anthropic waxay ku qiyaastay in wadarta dib u dhacyada nooc kasta ah ay ku dhici doonaan qiyaas ahaan 67% Claude.ai, 55% wada shaqaynta, 17% ee Claude Code, iyo 7% ee Claude Platform. Kuwani waa cabbirada shirkadda, ma aha natiijooyinka hantidhawrka madax-bannaan.

Habka ayaa sidoo kale muhiim ah: codsi calaamadeysan ayaa dib loo rogaa halkii uu ka jawaabi lahaa Fable 5. Taas oo ilaalinaysa helitaanka qaab guud iyada oo la ilaalinayo awoodda Anthropic waxay tixgelisaa inta ugu badan, laakiin ma caddaynayso in jawaab kasta oo caafimaad oo la oggol yahay ay sax tahay ama ku habboon tahay go'aanka bukaan-socodka.

For organizations, the practical question is how the boundary behaves across contexts. A student asking for a plain-language explanation, a clinician checking terminology, and a researcher requesting an experimental protocol may use overlapping words while presenting very different risk. Anthropic's routing approach can preserve a safer general answer for the first two cases, but only if the classifier recognizes intent, conversation history, and requested operational detail without turning a legitimate professional workflow into an opaque denial.

Maxaa la daawan doona xiga

U fiirso caddaynta in heerka dib-u-dhaca hoose uu la mid yahay ogaanshaha xooggan ee codsiyada dhabta ah ee khatarta ah, oo lagu daray xeerar cad oo gelitaanka cilmi-baarista la aamini karo.

Anthropic ma daabicin go'aanka qiimaynta, heerka beenta ah, ama ku celcelinta madax banaan ee ogeysiiskan. Shirkaddu waxay sheegtay in faa'iidooyinka beenta ah ay sii jiri doonaan iyo in kalasaarayaashu ay sidoo kale u adkeystaan ​​​​isku dayga jebinta, marka tirada 85% waxay cabbirtaa dib-u-dhacyo yar halkii ay ka ahaan lahayd ganacsiga buuxa ee badbaadada.

Shaacinta soo socota ee faa'iidada leh waxay tusi doontaa waxqabadka guud ahaan ereyada, luqadaha, wada sheekaysiga leexada badan, iyo socodka shaqada ee karti u leh qalabka. Cilmi-baarayaashu waxay sidoo kale u baahan yihiin inay ogaadaan inta jeer ee codsiga waxyeellada leh uu ka gudbo xadka cusub iyo sida ugu dhakhsaha badan ee kala-soocida loo cusboonaysiiyo marka marin-marin cusub ay soo baxaan.

Anthropic waxay sheegtay inay horumarinayso dariiqooyin la aamini karo ee awoodaha bayoolaji ee xadka. Kalsoonidooda waxay ku xirnaan doontaa cidda u qalanta, waxa la socodka iyo ilaalinta asturnaanta ay khuseyso, sida dhacdooyinka dib loogu eego, iyo haddii cilmi-baarayaasha sharciga ah ay ka doodi karaan xaddidaad khaldan.

The next disclosure should also explain how the classifier is evaluated after deployment. A lower fallback rate can be achieved by reducing false positives, by shifting difficult cases to another model, or by missing more harmful requests; those outcomes have very different safety meanings. Useful reporting would include false-positive and false-negative estimates by request type, performance under multi-turn escalation and paraphrase, language coverage, handling of tool calls, and the process for updating the boundary after a jailbreak or an incident.

The user-facing promise should be tested with the same care as the safety boundary. Anthropic says the update should help with everyday health, clinical, and educational questions, but a lower fallback rate does not establish medical accuracy, appropriate triage, or suitability for professional decisions. Independent reviewers should sample allowed answers for unsupported certainty, missing safety advice, and harmful procedural detail, while also checking whether the fallback model communicates its limits clearly. The strongest evidence would compare matched requests before and after the classifier change and publish enough anonymized examples for outside researchers to understand both the gains and the new failure modes.

That evidence should be reported separately for consumer chat, coding, agentic workflows, and the API because the same classifier boundary may carry different tools, context windows, and user expectations in each product. A single blended fallback percentage can hide a meaningful regression in one surface behind improvement in another. Publishing the denominator, confidence intervals, and product-level counts would make the result useful to educators, clinicians, developers, and safety researchers rather than only to readers comparing one headline number.

Tilmaamaha la xidhiidha & su'aalaha