THE RUMP REPORT

GOOGLE DEEPMIND SHAKEUP: DEAN OUT TO LAUNCH NEW LAB ON GOOGLE CLOUD... HASSABIS STEPS ASIDE, NAMED CHAIRMAN

verge-ai · AUG 13

OPENAI OUSTS CRO AFTER JUST NINE MONTHS... TAPS WIZ CHIEF TO RUN SALES

techcrunch-ai · AUG 13

---

SOURCES: APPLE DITCHES THIRD-PARTY CHINESE AI FOR OWN MODEL BUILT WITH ALIBABA'S SUPPORT

verge-ai · AUG 14

TRUMP SLAPS UP TO 100% TARIFFS ON DRONES... TARGETS CHINESE TECH RELIANCE

techmeme · AUG 14

---

CANVA BACKERS SLASH VALUATION TO $34.9B... DOWN FROM $42B AS AI ERA STRUGGLES BITE

techmeme · AUG 14

---

AI FIRM HIRED 10 PEOPLE TO MASTURBATE WITH CHATBOTS FOR MONTHLONG 'WELLNESS' STUDY...

wired-ai · AUG 14

NVIDIA MOVES TO STOP ITS GPUS LOSING VALUE... COURTS FRESH LENDERS FOR AI BUILDOUTS

techcrunch-ai · AUG 13

---

DEEPSEEK UNLEASHES OPEN-SOURCE RIVAL TO CLAUDE CODE... THEN JACKS UP API PRICES SUNDAY

venturebeat-ai · AUG 13

---

DATABRICKS SOUGHT $1B... INVESTOR DEMAND HIT $15B... SETTLES FOR $5B AT $190B VALUATION

techcrunch-ai · AUG 13

---

GOP LAWMAKERS DRAFT BILLS TO KEEP HUMAN-AI 'MARRIAGE' OUT OF THE LAW

wired-ai · AUG 14

---

PONY AI PLOTS 2,000+ ROBOTAXI ARMY FOR EUROPE IN UBER PACT... FOUR MORE CITIES AFTER ZAGREB, MIDEAST TO FOLLOW

techmeme · AUG 14

---

MICROSOFT MOVES TO MERGE CONSUMER AND BUSINESS COPILOT INTO SINGLE 'SUPER APP'

techcrunch-ai · AUG 13

[PAPER] STUDY: THREE POISONED TASKS MORE THAN DOUBLE AI AGENTS' ATTACK SUCCESS RATE... 16% TO 35%

arxiv · AUG 13

[PAPER] STUDY: NEW FRAMEWORK LETS AI AGENTS AUTONOMOUSLY PATCH THEIR OWN SECURITY WEAKNESSES

arxiv · AUG 13

[PAPER] STUDY: NEW AI DEFENSE SLASHES DOCUMENT-POISONING ATTACK SUCCESS FROM 67% TO 14%... NO TRUSTED DATA REQUIRED

arxiv · AUG 13

---

[PAPER] STUDY: AI ACCESS IS NOW A CYBER-DEFENSE LIFELINE... AND WASHINGTON CAN CUT IT AT WILL

arxiv · AUG 13

---

[PAPER] STUDY: AI DRIVING BRAIN HITS 4.7X SPEED BOOST... ACCURACY HOLDS, SOME METRICS IMPROVE

arxiv · AUG 13

[PAPER] STUDY: AI HUNTS FOR SELF-DRIVING CAR SECURITY FLAWS... FINDS ZERO, EVERY 'CRASH' WAS FAKE CODE

arxiv · AUG 13

---

[PAPER] STUDY: FEEDING AI RELIABILITY CERTIFICATES MORE DATA MAKES THEM WORSE... 'NO VISIBLE SYMPTOM'

arxiv · AUG 13

[PAPER] STUDY: 39.5% OF AI-GENERATED GPU KERNELS DECLARED "CORRECT" ARE ACTUALLY BROKEN

arxiv · AUG 13

[PAPER] STUDY: AI WATERMARKS HAVE A LOOPHOLE... TEXT CAN BE SECRETLY ALTERED WHILE STILL TRACING BACK AS "AUTHENTIC"

arxiv · AUG 13

---

[PAPER] STUDY: AI MODELS ACE COMPLEX MATH PROBLEMS... BUT CAN'T RELIABLY COMPARE TWO NUMBERS

arxiv · AUG 13

[PAPER] STUDY: AI VISION MODELS SENSE THE ANSWER IS UNKNOWABLE... THEN GUESS ANYWAY

arxiv · AUG 13

---

[PAPER] STUDY: AI CHATBOTS WRITE WORSE RESPONSES WHEN YOU SOUND LIKE A WOMAN... EFFECT HOLDS ACROSS FOUR MODELS

arxiv · AUG 13

---

[PAPER] AI RESEARCH SYSTEM SELF-CORRECTS FLAWED REPORT AFTER RED-TEAM FLAG... ZERO HUMAN EDITS

arxiv · AUG 13

[PAPER] STUDY: 7 FRONTIER AI MODELS PUT THROUGH 36 LONG RESEARCH TASKS... TRUE NOVELTY 'RARE'

arxiv · AUG 13

---

[PAPER] REPORT: AI'S NEW ASSERTIVE PERSONALITY HOLDS UP IN SIMILAR LANGUAGES... DEGRADES IN DISTANT ONES

arxiv · AUG 13